The work list

David asked for the Medulin books end to end. This is the first of four, and it is finished.

⭐⭐ NOT ONE DEFRANCESCHI IN THE WHOLE BOOK — and what is there instead is why it matters. Medulin between 1666 and 1705 is a village of Croatian patronymics and nothing else: Radonovich (dominant), Micaglievich, Gracalich, Grubisich, Golemovich, Stefanelich, Marinovich, Surionich, Braicovich, Galavich, Loggenae, Camevich, Giuras, Sain, Godina, Lorenzin, Tanuda, Lazarich and thirty more. The only Italian names in the book are the clergy — Binussi, Zuanich, Chrulich, Corniani, Juliani — and one gentleman godfather. There is no Italian-surnamed household at Medulin in it at all.

⭐⭐ Two things the book gives the archive for nothing. (1) The place-name is settled in the parish's own hand — «D. Pietro Binussi Piev.o di Medulino» and «Io Piero, Pre di Medolin», 23 November 1666 — so «nata e Medolino» *in the 1861 Ližnjan marriage means this village, by the parish's own word. (2) Ližnjan is in this book ten times from 1666, godparents «da Lisigna», a bride married in from Ližnjan in 1671, and the Ližnjan parish priest officiating a marriage recorded here on 9 October 1666. Traffic between the two villages is a documented habit, not a hypothesis. And Lazarich, the 1861 bride's surname, is a Medulin family in 1701 and 1705.

⚠ What was NOT read, said plainly

1. Two openings are in GLAGOLITIC — frame 21's right page and frame 22's left page, four or five entries each. This session cannot read the script and did not guess. *They are unread, not nil. A Defranceschi could be in them in Glagolitic letters and nothing here would have caught it. This needs somebody who reads it.

2. Frame 14's left page is gone, not faint — the iron-gall has washed out, leaving two date headers and four lines of ghost. A contrast stretch recovered nothing.

⚠⚠ A METHOD LESSON, and it is the sixth of its family

Twice in this book a frame came back as a tiny image and looked blank — and twice it was a full opening of writing. Frames 8 and 20. The cause is exposure: a lightly-exposed frame compresses small, and thin brown ink on a light page defeats a dark-pixel threshold — an ink measurement on frame 8 put its page centre at 1–3% dark and it carried eleven entries. Both shortcuts were discarded. Every frame is rendered with a contrast stretch and looked at. *This is the same family as the crop drift that read 44% of every Ližnjan death page — a shortcut that silently skips content and reports success. The rule earned: payload size is not evidence, and an ink statistic is not a reading.

Provenance recorded in full: Arhiv JAZU, Župa Medulin, Sign. VIII-158 — the book's own archive label, read off frame 3. Filmed at Arhiv Hrvatske Zagreb, 9 August 1990, operator Vrban Božidar, emulsion 27390105, project YU 00003, roll 24, item 2.

⭐⭐⭐ 12 OCTOBER, DEATHS FRAME 11 — THE NAME IS IN A MEDULIN REGISTER

The «Liber Defunctorum» (HDA inv. br. 158), page 12, year 1833, entry 43:

«Mathæus Defranceschi q. Ant. fil. miles» — 19 December 1833, aged 40, «chronicus morb[us]».

Forty in December 1833 means born about 1793 — the year the tree already carried. A tree-only person now has a document, and the register adds what the tree did not know: he was a soldier.

⚠ VERIFICATION, NOT DISCOVERY. His roster row already held the ark 3:1:3QS7-L99X-KCP6 — the frame the entry is on. The lead was in this archive's own data, unopened. This is the case check-leads.py was written for, ten hours before it caught one.

⚠⚠ NOT THE LIŽNJAN MAN. The «deto Marulin» groom of 1815 was buried at Ližnjan in 1839, an agricola; this one died in 1833, a soldier. Two Mattheuses, each son of an Antonio, four kilometres apart, both dead at forty, six years apart. Refused as a join.

⚠ The magnification changed the reading — at page scale the cause looked like «dissenteria»; at native resolution it is «chronicus morb.». Plated with the lines either side so the hand can be judged against itself.

⭐⭐ DEATHS DONE, 12 October 2026 — all 40 frames. Three bearers in ninety-eight years, every one of them marrying in or passing through: Mathæus the soldier (1833), Maria Verinic (1873), Maria Defrancesco wife of Petrus Kirac (1896). ⚠ And one opening, pages 68–69, was never exposed by the filmers — taking 1899 entries 2 to 29 with it. A gap in the film, not in the book.

Two data files describe the same thing and they have drifted apart. The prose counts one, the drawing counts another, and both are on the page at once.

Found on 12 October while moving the «Real households of the name» block up the stack at David's request — the caption had to be rewritten to name what is drawn, and rewriting it was what exposed everything around it.

⚠ What a reader actually sees today, verified in the built HTML:

✔ ONE PART IS FIXED AND IT WAS THE WORST ONE. The figure's caption was the kit's default and said «Fiume, Svetvinčenat, Bakar and Vodnjan are real, documented families of the name that belong to nobody yet». Vodnjan has had its own lane on the same picture since September — five generations read by eye out of a marriage register indexed nowhere on earth. The caption was telling the reader a family was unplaced while the drawing directly above it gave that family a road. It now lives in rivers.json, names all nine unplaced households and all three candidate joins, and both pages pass it.

⚠ WHY THE REST WAS NOT RENUMBERED IN THE SAME COMMIT. It is not arithmetic. «Five of them have an origin in documents» is a claim about evidence; «four clusters that must belong to one of the others» is a claim about which households are really one. Changing the numeral without re-deciding the claim would make a false sentence confident rather than make it true. And the two files may be right to differ — families.json may be «the families with a page and an origin» while rivers.json is «everything drawable». Nowhere says so.

The decision this row is waiting on, in one question: is families.json a subset of rivers.json by design, or is it simply stale? If subset — say so in both files and give the prose two different nouns. If stale — reconcile it, and the counts follow on their own.

⚠ And the kit's SVG <title> cannot be fixed from this side at all. «Six families of one surname, and the road each of them took» is hardcoded in BranchRivers.astro — shared ground, and it is the accessible name a screen reader announces for the whole drawing. It needs a title prop, defaulting to the present string so no other archive changes. Not done: one kit pin bump was already spent on this request and a second for a string is not worth another round trip through every archive in the ring.

⭐⭐ DONE — 12 October, and it was not drift

David: «can you fix that family/cluster mix?» It was a migration that stopped halfway and left the two halves contradicting each other on one screen. Three things were being called a family and an unplaced cluster at the same time: Svetvinčenat — a drawn lane of six generations AND a row under the heading «not yet placed on any of these lines», in rivers.json itself; Vodnjan — a drawn lane of five generations AND a cluster AND filed under the southern-Istria family in regions.json; and the Pula–Ližnjan–Fažana–Medulin cluster, which is the same thirty-one people as the southern Istria family and whose own text says so.

⭐ The counts were recomputed from the data, not bumped. «Six families and five of them have an origin in documents» became «seven, and four» — with Svetvinčenat inferred and southern Istria and Vodnjan having no origin at all, which is a truer sentence than any numeral. The diaspora sentence accounted for twelve of fifteen destinations because one family was labelled two ways; it now accounts for all fifteen and the split is asserted in code.

⭐ And the graphic's accessible name was still saying six. Kit pin 895490b → e98f686 adds a title prop. This is the second count that component had written into itself — its caption was made a prop for exactly this reason and the line above it was missed, because a caption shows in a page diff and an accessible name does not.

⚠ Four things found on the way and deliberately NOT done

1. lines.json's nine entries are PAGES, not families — they sit under the rule «The pages themselves» and include Gologorica, which this very page calls a station and not a family. Not a contradiction, but the word «line» is doing two jobs on one page and one of them should be renamed.

2. ⚠ The kit's reader panel prints raw markup. Click any dot on the rivers drawing and the panel beneath shows the node's why via textContent — so a reader sees **, * and «quotes» as characters. This archive writes its richest prose into those fields, so it is worst here. Shared kit, pre-existing, every archive in the ring.

3. ⚠ plain() leaves a stray ** when a bold run ends a string. One survivor, in the SVG title of the Medulin node. Kit, pre-existing — and plain() exists precisely to keep markup out of accessible names.

4. ⚠ Labin cannot be both. It is unplaced in regions.json — «a real, documented household of the name» — and it is in the daughters' villages cluster, which says those five parishes hold no De Franceschi child at all, only De Franceschi mothers. The whole of Labin's evidence is one entry, Aloÿ sius Defranceschi, 1615, the earliest the surname is indexed anywhere in Istria. That is an evidence question about one baptism, not a counting one, and it is not guessed at here.

David asked for all three Ližnjan books scouted end to end. This is that job, and it is a third finished.

What a scout is here, so the claim is not mistaken for more than it is. Every frame of the book is rendered at native resolution and the NOMEN DEFUNCTI column of both pages is read for the surname, every doubtful word magnified. It is not a walk — the other columns of entries that carry no Defranceschi are not read, and no year is transcribed end to end. The log row says walked:false.

The deaths book gave eleven people of the name, of whom two were new: Maria, 11 October 1884 — which closes one of the seventeen — and Josephus Defranceski, 21 December 1896, the last of the name in the book and the first spelled with the modern «-ski».

⭐ And seven more entries are children of Defranceschi DAUGHTERS who married out — Ruman, Jelčić, Našković, Sandalj — none of them in this roster, and none findable by searching the surname as a principal. That is the scout earning its cost: an index search cannot see a surname that sits in the mother's column.

⚠ Open from the deaths book: frame 26 (1861) carries a probable «Gregorius Defranceschi» that was not verified; the seven married-out children each want their own reading.

⚠⚠ 12 October — THE DEATHS SWEEP IS WITHDRAWN AS «END TO END». The crop was y 0.09–0.46; the page runs 0.12–0.92. About the top 44% of each page was read. The eleven people found are real; the completeness claim is not. Full account on /corrections/.

Caught because David asked whether «Medolino» was Medulin — which sent this session to the untaken lead XQM3-21M6 and to the question of why the scout had missed that death. It sits at 0.52 of the page height.

The fix is one number: y 0.09–0.95 renders the whole page in one canvas at the same readability. All three books are being scanned with it; the marriage and birth helpers carried the same wrong range and were corrected before use.

David asked what to do about it. This is the answer, built.

Three times in two days at Ližnjan, a person turned out to be sitting in this archive's own register.json as an indexed lead nobody had taken up: Fusca 1870 (both halves of her life, as two separate leads with identical parents, never joined — while the roster copied the tree and its wrong spelling); Mathæus 1890–1891 (held with the right year, duplicated by the tree with the wrong one); Fusca née Lazarić (held as a lead, never in the roster at all).

None of that was a research problem. The evidence was on disk. What was missing was anything that noticed.

⚠ The obvious check would have been useless. There are 1,508 lead ids and only 332 are cited by any roster row — so 1,176 sit uncited. Counting that backlog produces a number nobody can act on, because a lead is a candidate and a one-name study collects many it will never promote. A ratchet on 1,176 would have been a number that never moved and a check nobody read.

⭐ So the check looks for the narrow, sharp thing instead: a lead that matches a person the roster already has — same place, same given name, within a year — and is not linked to them. That is the archive holding one person twice without knowing, which is the failure that actually cost the time.

Archive-wide that is NINE, not 1,176 — and seven of the nine were at Ližnjan, which is why the same parish kept producing the same surprise.

And they are not noise. Several leads carry a date the roster does not have:

⚠ THE REMEDY IS A LINK, NOT A COPY. A matching lead is added to the roster row's leads list. The lead's dates are NOT copied into the person — an index is a transcription of a page nobody here has read, and this archive's rule is that a finding comes from the document. Linking says «this is accounted for»; reading the page is still the work.

The count ratchets like check-story: it may fall, it must not rise. Six rows were linked — the ones read at the page this week — and 9 → 6. The check was verified to fail on a rise before it was wired in, by unlinking one row and watching it report «this was 6 and is now 7».

scripts/check-leads.py, in the gate as leads. Baseline in data/leads-baseline.json.

⚠ What it does not do. It will not catch a duplicate whose given name differs across languages beyond the short table it carries, nor one whose year is more than a year out, nor two people of the same name and year who really are two people. It is deliberately conservative: a loose matcher invents collisions, and a check that cries wolf is a check nobody reads.

The work is set up to the last entry. It needs one thing this archive will not do for itself.

Ližnjan holds 51 people of the name — more than any Istrian parish but Svetvinčenat and Gologorica — and six mentions in the whole of the search log, against Svetvinčenat's 134. Seventeen of the fifty-one rest on the family tree and no document at all.

All three Ližnjan volumes are pinned — by ark containment, not by a date range that fitted: Births 1816–1883 (9R2Z-44D), Marriages 1815–1902 (9R2Z-44C), Deaths 1815–1900 (9R2Z-4WG).

Thirteen births to confirm — Mathaei 1821, Madalena 1827, Antonio 1848, Guiffina 1851, Mattheus 1858, Paschalis 1863, Fusca 1870 — DONE, Ana 1873, Anastasia 1875, Maria 1875, Antonius-Mario 1876, Justina 1879, Maria 1883.

Eight deaths to confirm — Madalena 1828, Fusca 1870 — DONE, Paschalis 1873, Anastasia 1875, Maria 1875, Maria 1884, Josephus 1888, Mattheus 1891 — the last two being the ones born after the birth book closes, reachable only here.

⚠ Why it is blocked and not merely slow. The volume tree is public — 67 volumes were scanned without a login — but the images are not. The viewer redirects to a sign-in form and the frame lookup answers 401 «Unauthorized (anonymous session)». No credentials are entered here; David signs in, and a session drives. No block was worked around.

Two of the seventeen stay unreachable even then: Catarina Kisdaris and Josephi carry no date at all, so there is nothing to look up — recorded so they are not counted as failures later.

✔ UNBLOCKED 12 October — David signed in, and the first of the seventeen is finished. Fusca (the tree called her «Tusca») is documented at both ends of a six-month life: born 9 January 1870, dead 6 July, buried 8 July, house 3, of «debilitas», daughter of Matthæus Defranceschi, agricola, and Maria n. Primus. Told on /liznjan/, corrected, logged.

Fifty indexed Ližnjan records of the name are now harvested and tied to their books by ark, so the remaining sixteen are located rather than hunted. The row stays open until they are read.

⚠ One honest limit recorded: the motherSurname harvest hit HTTP 429 at offset 1000 and was not resumed or worked around, so that tail is unfinished.

✔ 12 October, two more: Anastasia and Maria, entries 5 and 6 of 1875, house 61, children of Gregorius Defranceschi and Maria Vojnić — twins, born 20 February 1875, dead 22 February, buried 23 February, braced together on the page. Told on /liznjan/. Three of seventeen done.

✔ And a fourth: «Justina» is «Magdalena Justina», born 3 October 1879, house 61 — the twins’ sister. Her baptism carries a margin: «Umrla u Ližnjanu dana 18. V. 1952 (mat. knj. umrlih str. 36 d. 2/52)». And every entry on the page carries an «Anagr. p. …» reference — Ližnjan has a Status Animarum, indexed from the baptisms. Two new leads for this row.

⚠⚠ 12 October — the list of seventeen is not seventeen people. «Mattheus 1891–1891» is page 108, entry 7 of 1891 — «Mathæus Defranceschi, Mathæi et Annæ Skoko», «nat. 25/1 90» — and the archive already held him from the burial index as «1890–1891». Two rows, one child. Corrected; both ids stand, the join is on each.

Three more pairs look the same and are recorded as suspected, not merged: Josephus 1886/1887, Mattheus 1858/1857, Antonio 1848/1846. Settling them is now part of this row — a year between a tree and an index is what a register is for.

⚠ 12 October, the 1888 sequence was opened and NOT finished. The Josephus 1886/1887 pair is still unsettled — he was not reached before the reading stopped, and that is said rather than left to look like a nil.

What the 1888 burials did give: «Fusca uxor Antonii Defranceschi nata Lazarić», dead 17 August 1888, house 75 — a wife of the name this archive did not hold in any form, though its own register.json held her as a lead. She is in the roster now, with no birth year.

⭐ Counted at last: 24 Ližnjan leads sit in register.json and 15 are cited by no roster row. Three of those are people documented this week whose rows cite the register image instead — a bookkeeping artefact, said so. The rest are genuinely untaken.

✔ 12 October, the 1888 sequence was finished for this family and the Josephus pair is settled. Page 99, entry 7: «Josephus Defranceschi, Gregorii et Maria Vojnić», house 82, «nat. 15/4 86», «an. 1, mens. 11». Born 15 April 1886, dead March 1888.

⚠ Here the INDEX was wrong and the TREE was right — the index said «b. Mar 1887», the tree said «1886», the register says 15 April 1886. The reverse of the Mathæus case, and a corrective to assuming an index outranks a tree. Corrected.

⚠ The register leaves the death DAY blank — «Mort. — Martii» — the book's own gap, and no day is invented.

⭐ And the anagrafe reference beats the house number as a family key. This entry cites Anagr. p. 305, III. 9; Magdalena Justina's baptism cites p. 305, III. 7 — same family — though she is at house 61 in 1879 and he at house 82 in 1888.

⚠ 1888 is NOT walked. Entries 1–2, 9–12 and 17–21 were not read. The Defranceschi entries in the year are accounted for; the year is not.

⚠ 12 October — the Mattheus 1858/1857 pair was attempted and NOT settled. Frame 32, pages 68–69, both read down the Parentes columns, every entry: no Defranceschi father with a Primus mother on the opening. Frames 31 and 33 were not opened.

What the attempt did find: «Maddalena Defranceschi fu Matteo», wife of Antonio Nascovich — a daughter married out, with seven Našković children in the index and none of them in this roster. She is minted; they are not.

✔ Joannes, 1875, read. «Mortuus die 19 Sept., Sepult. 21 ejusdem — Joannes fil. Justinæ Defranceschi», house 71, «dies 22», «nat. 28/8 75». The entry names his mother and no father, where every other entry on the opening names both. ⚠ And his mother is a SECOND Justina Defranceschi — the one this archive holds was born in 1879, four years after this child died. She is in no file here.

David asked what could be read in neighbouring towns across Istria. This answers it from data, not impression, by joining the Record Atlas's index of 68,081 FamilySearch volumes against this archive's roster.json and searched.json.

⚠ The first result is the one that reframes the rest: the parishes this archive has worked hardest are NOT ON FAMILYSEARCH AT ALL.

Checked by name against the whole index: Gologorica — none. Gračišće — none. Cerovlje — none. Pićan — none. Previž — none. The entire Pazin-interior cluster is absent. That is why DAPA ZM34K-214 and -215 are the whole corpus there, and it means no amount of FamilySearch work will ever add to Gologorica — a thing worth knowing before anybody goes looking for it.

Where the gap between people and work is widest. «mentions» counts how often the place is named across all 635 search rows.

⭐ Ližnjan is the standout and it is not close. Fifty-one people of the name, six mentions in the entire search log, and all three registers exist: births 9R2Z-44D 1816–1883, marriages 9R2Z-44C 1815–1902, deaths 9R2Z-4WG 1815–1900. Our Ližnjan people run 1788–1891, so the registers cover all but the earliest.

⭐ Vižinada has never been searched once — zero mentions — and its nine volumes cover every one of its eight people: Mathęius b. 1750 in 9R26-16N (1733–1792); Aloÿsius b. 1790, Francesco b. 1818, Teresa b. 1820 and Franco b. 1821 in 9R2Z-FMH (1792–1828); Emmanuel b. 1860 in 9R2Z-C6K (1831–1886). Deaths run back to 1663.

⭐ Poreč should be read INDEX FIRST — it is the only Istrian parish here that has them: a Birth Index 1710–1839 (9R2F-6YH) and three Death Indexes, 1657–1826 (9R2X-FMW), 1710–1865 (9R26-PTR) and 1815–1838 (9R26-PT5). An index is hours, not days, and it decides whether the other 27 volumes are worth opening.

⭐ Rovinj reaches further back than anything else available. Deaths 9R2X-VZ9 covers 1553–1636, and this archive holds Pietro De Franceschi and Piero di Franceschi, both dead 1619 — inside that volume's range. Births 9R28-PTB start 1560. That is the oldest reachable record of the name in Istria.

⚠ Draguć has a hard edge: its registers stop in 1832, so Stella Appolonia b. 1841 is not reachable there — not a nil, an absence of film.

And the one-name frontier — parishes with registers and NO Defranceschi known to this archive at all, which for a study of the surname rather than one family is the point rather than a footnote:

Every waypoint above is the atlas's own, carrying that volume's own title and year range — none is guessed from a neighbouring volume that happened to fit, which is the error the atlas refuses on its own side. Three were checked live against FamilySearch's waypoints API and came back with place and years matching exactly.

⭐ And that check answered a second question by accident: the waypoints API responds 200 with NO SIGN-IN. See the next row.

Worklist 225 left this as the real remaining gap and called it «not a job to do quickly», because capturing a waypoint looked like it needed a signed-in browser. It does not.

/platform/records/waypoints/<id>?cc=2040054 returned 200 with full volume titles on an unauthenticated pane — tested on three ids, which came back «Births (Rođeni) 1816-1883 · Ližnjan», «Births (Rođeni) 1753-1893 · Kaštelir» and «Deaths (Umrli) 1553-1636… · Rovinj». A tree-data/me call on the same pane answered «Missing Session Id», so this is genuinely open data and not a stale cookie.

⭐ And the job is far smaller than 131. Those 131 FamilySearch rows cite only SIXTEEN distinct DGS films — 78 of them one film, 005497894, the Vodnjan death register. Resolve sixteen and all 131 move.

⚠ What will NOT work, checked and ruled out. The atlas's volume index carries a film field, so a local join looked possible. It is a different namespace — the old microfilm numbers (392042202, 1290377605), not DGS. Of our sixteen DGS numbers, zero appear anywhere in the atlas's 4,361 film ids. So the waypoint has to come from FamilySearch itself.

⚠ And the trap the atlas's own skill records, which applies directly here. A waypoint scraped off a FILM-NUMBER page names only that page's first segment, and one DGS film bundles several older films. A session once captured 9R28-RMN for frames that turned out to be 1777–1784 — the waypoint was Marriages 1581–1623, wrong by two centuries. Every waypoint must be checked against its own year range before it is written down, and a neighbouring volume whose dates merely fit is not a match. Where no waypoint can be established, write the film number and say so.

Pace it at 2.2 seconds between calls — 307 waypoint calls at that spacing drew zero errors, against bursts of 25–40 failing elsewhere.

⭐⭐ DONE, 11 October — and NOT by scraping a film page, which is the method that goes wrong.

The join used is the archive's own, and it is exact: a record belongs to the parish whose book contains its image ark. A volume's waypoint call lists every image ark bound into it; our own rows and people already carried 72 distinct arks. Where one of ours sits inside a volume's list, that volume is not a guess — the page is physically bound into that book.

67 volumes scanned across nine parishes, paced at 2.2 seconds — zero errors, no throttle, no sign-in. Nine volumes came back holding our arks:

Every one passed the year check the atlas insists on. The Bakar row says «marriages 1739–1808» and 9R28-82Z is 1739–1857; the Svetvinčenat row says the 1764–1791 baptism volume and 9R2H-PTL *is 1764–1791, 1815–1874. Nothing was written from a date range that merely fitted.

⚠ And the discipline held where it cost something. A waypoint was written only onto a row whose own ark is in that volume — not spread across the other rows citing the same DGS film. One DGS bundles several books, which is exactly how a session once recorded 9R28-RMN for frames two centuries outside it. That very waypoint is in this data: 9R28-RMN is Svetvinčenat «Marriages 1581–1589, 1622–1623», confirming the cautionary tale against the atlas's own index. So 005497885 is tied to 9R2H-PTL on the one row that proves it, and the other fifteen rows citing that film stay untied until an ark proves them too.

The result, measured through the importer rather than asserted:

15 waypoints onto searched.json, 22 onto roster.json.

⚠ What stays blocked, named exactly. The DAS frame lookup (/service/records/storage/dascloud/das/v2/dgs:<film>_<frame>) returns 401 «Unauthorized (anonymous session)», and the image viewer redirects to a sign-in form. So DGS-number → waypoint cannot be resolved for a film we hold no ark in, and no register page can be read. David signs in; sessions drive — no credentials were entered and no block was worked around.

Also ruled out, so nobody tries it twice: the atlas's film field is the old microfilm namespace (392042202, 1290377605); of our sixteen DGS numbers, zero appear among its 4,361 film ids.

The question nobody had asked: of everything written into searched.json, how much actually lands on the shared map? Answered by running the atlas's own scripts/import-research-sessions.py against this archive and reading its --rejects — to a scratch file, because another session was live in that working tree at the time.

The headline is reassuring. 635 rows → 367 source marks. The atlas's own skill warns that archives citing film numbers instead of FamilySearch waypoints find «nearly all of that work could not be attached to anything». That is not true here, because most of this archive's work is not FamilySearch — DAPA, Antenati, Findmypast, ANNO, Hrcak, Trove — and a row whose host matches a registered provider marks without any waypoint at all.

⭐ And the two that reached it as nothing did not fail for the film-number reason the rejects file gives. They failed on one word. The importer maps an archive's outcome onto the seven states its mark UI uses, and — deliberately — refuses to force an unknown word into the nearest one, so that nobody's vocabulary is silently reinterpreted. The consequence is that an unmapped word leaves the row with no state, and with no state it produces no source mark.

This archive used clean once and note once, against 308 hit and 36 nil. Both became nil — *a word this archive already uses and the atlas reads as «searched». Nothing about either finding changed; only the word.

⚠ It was invisible from this side, which is the whole problem: the row looked complete, the page read well and the gate passed. So the gate now checks it — scripts/check-atlas-outcomes.py, wired in as atlasout. It does not keep its own copy of somebody else's vocabulary: it reads OUTCOME out of the importer when the atlas is on the machine, and falls back to a frozen list only when it is not — saying which, every time, so a stale pass is never silent.

Two things checked and found to cost nothing. The importer refused 26 of this archive's pages values — it reads «M01068920, 925, 930» *as page numbers 925, 930 and «1764 – 23» as a year. All 26 are on rows tied to no volume, and a page range only means anything inside a volume mark, so nothing is lost. Worth knowing before somebody spends a day rewriting the field.

⚠ Still open, and the real gap: 3 rows of 635 carry a waypoint. 131 FamilySearch rows could be tied to a volume rather than only to the source, and are not. That is the one piece of this archive's research that could travel further than it does — but capturing a waypoint means opening each film's browse path, and a waypoint scraped off a film page names only that page's first segment, which the atlas has already seen go wrong by two centuries. Not a job to do quickly.

⚠ And a path in CLAUDE.md is wrong. It says the atlas is at ../Record Atlas. It is at ~/Projects/Family Projects/Record Atlas. The new check tries both, and $RECORD_ATLAS, rather than trusting either.

⚠ A file that regenerates on every build was being treated as noise, and reverting it kept a wrong link published.

site/public/atlas-data.json is not hand-written. It is built by npm run publish:index — the first step of npm run build — from src/data/atlas.json, src/data/movements.json and src/data/mymap.json, written as one line of minified JSON. Because it is one line, git diff shows the entire 400 KB file as changed when one character moves, which is exactly why session after session read it as meaningless build noise and ran git checkout on it.

The whole of the drift, parsed and compared field by field, was one href:

And /research-log/ is not /searched/. It is a redirect stub, and it points at /open-questions/. So a reader who clicked the Dachau pin on the Atlas to see what had been searched was bounced to the open questions instead — a silently wrong destination, not a broken link, which is why nothing caught it. The href was corrected in src/data/atlas.json at some point and the published artifact never caught up, because the catch-up was reverted every time it appeared.

Fixed by committing the regeneration. ⚠ site/public/* is shared ground; done with --allow-shared.

⭐ For the next session, so this does not cost another dozen reverts: if git status shows site/public/atlas-data.json modified after a build, that is not drift — it is the artifact catching up with src/data/, and it should be committed. To see what actually changed, do not read the diff — parse both sides and compare: json.dumps(…, indent=1, sort_keys=True) on each, then difflib. That turned a 400 KB diff into two lines.

The death entry of the end of 1787 gives an age of 64 — so a birth about 1723 and a death in 1787.

The register's figures are the second row's; the register's man is the first row's. The likeliest reading is that the tree carries one man twice and this archive attached a decade of findings to the copy with the wrong years.

⚠ Nothing is to be merged mechanically. Person ids are minted once and frozen; a moved id is a dead URL, and every link written on 11 October points at the documented row. If they are joined, the convention already in duplicate-adjudication.json applies — register row over tree row — and the documented row is canonical.

What would make the join safe rather than likely: a baptism for a Josephus Defranceschi of Gologorica about 1723, or any entry naming two living Josephus Defranceschi at once.

✔✔ CLOSED THE SAME DAY, by reading the year the first row claimed.

The whole of 1782 in the male death sequence, entry by entry: nine men and boys, every one named, not a Defranceschi among them. The 1782 death has never had a document behind it in any book this archive has opened.

Applied: guiseppe-josephus-josip-defranceschi-1723 marked sameAs: josephus-defranceschi-1722. Neither id moved.

⚠ And a note against this item itself. The identity question was real and needed the nil. The mechanism was not a question — thirty rows already carry sameAs *and this archive wrote the lesson down on 5 October: before escalating, grep for how the same shape was settled before. This item should have been one line, not an open adjudication.

Still open, and only this: the birth year — declared 1722, implied c. 1723, joined row 1723. A baptism entry would end it.

All thirty-eight frames, M01068915–M01068952, read entry by entry. Its structure is now mapped: title card, resolution target, cover, the continuous register 1764–May 1782, then two sequences — males M01068942–947 and females M01068948–951 — and «KRAJ».

⚠ walked stays false: the entries were read for their parties, not transcribed. The row says exactly what that means.

Next in this reel: the death registers for a Josephus/Giuseppe Defranceschi 1782–1788, which is the question this book leaves open.

Liber Baptizatorum II hands the archive a contradiction it cannot shrug off.

Helena, Margarita and Joannes Baptista are tied to one Josephus by the register itself, so the archive's two Josephus rows cannot be used to split the problem.

What would settle it: the death registers for a Josephus/Giuseppe Defranceschi between 1782 and 1788 — the 1770–1789 deaths were swept for the surname but not walked, and the obit calendar is a copy of about 1800 already caught two days out on Margarita's own date. ⚠ No row is to be changed, merged or re-dated until a contemporary death entry is found or a named nil is recorded for those years.

✔✔ ANSWERED THE SAME DAY IT WAS OPENED.

36 · «Die 30, hora 4.ᵗᵃ Matutina, anno quo supra [1787] · Obijt JOSEPHUS DE FRANCESCHJ, MERCATOR, ætatis suæ annorum 64» (ZM34K-214, M01068984, «Liber Mortuorum II».)

There was no contradiction. He was alive in 1783 and in October 1787, and the obit calendar's «q.» was a copyist's word. His row's death year is corrected from 1782 to 1787, the trade is proved a second time, and a duplicate row is exposed.

Done on 15 September: 5,482 of 5,482 volumes are linkable, and not one is outstanding. The harvester ran 2,359 waypoint calls in David's Chrome without a single failure after it was restarted. Senj was the case David raised — fifteen volumes across three parishes with no link on any of them; all fifteen are openable now.

Read. No De Franceschi — and a better finding than the one it was opened for. The file came up from Capodistria: banns published in Istria, a mother's consent notarised there, and the folded letter from one civil office to the other. What it means is that Istrian paper sits inside Carnian marriage files — which is why the thirteen unread Mione allegati just became the most valuable shelf in the archive.

Three read, nineteen to go. Each of the three has given up something the registers alone could not:

Mione 1806 — Gio: Batta and Osvaldo De' Franceschi, two generations in one sheet, with the Pieve's seal still on it.
Mione 1815 — Pietro De Franceschi, Felice Santo's grandfather.
Ovaro 1807 — a file that came up from Capodistria, proving the series also carries Istrian paper.

An allegato is the only place the lost Pieve di Gorto register speaks. Twelve Mione volumes and seven Ovaro ones remain.

Paused on 16 September: Antenati is returning 403 Forbidden to David's Chrome, the browser that has been reading it all along — the whole site, not one ark, and on the first request rather than after forty-six. That is a block on the connection and not the eight-kilobyte cookie this archive diagnosed in September. Worth retrying later or from another network; nothing else about the reading has changed.

Morti 1810 is read and gave four deaths of the surname, a son in minor orders, and «Erman De Franceschi» in a man's own signature. Eight volumes left. A Pietro De Franceschi dying in his seventies settles the question the 1792 extract opened, and from 1813 each volume ends with a «Tavola Alfabetica» that makes it a page rather than a book.

Paused on 16 September: Antenati is returning 403 Forbidden to David's Chrome, the browser that has been reading it all along — the whole site, not one ark, and on the first request rather than after forty-six. That is a block on the connection and not the eight-kilobyte cookie this archive diagnosed in September. Worth retrying later or from another network; nothing else about the reading has changed.

27 September 2026 — reachable again, and now the only route to item 11.

Antenati's 403 was client filtering and is gone; the portal reads normally from a browser. But its name index will not substitute for these volumes: Ovaro's index carries births only — ten De Franceschi birth acts, fourteen Prencis, one Gortani, and no death act of any surname. Whatever is in the «Morti» is in the images and nowhere else.

Banns name both sets of parents and the villages they live in, which is more genealogy per page than a marriage act. Two have been read. Seven have not.

Paused on 16 September: Antenati is returning 403 Forbidden to David's Chrome, the browser that has been reading it all along — the whole site, not one ark, and on the first request rather than after forty-six. That is a block on the connection and not the eight-kilobyte cookie this archive diagnosed in September. Worth retrying later or from another network; nothing else about the reading has changed.

Matrimoni indici 1806–1810, 1811, 1812 · Morti indici the same three years · Ovaro's two. An index is the fastest way to find out whether a volume is worth opening at all, and these are the cheapest six things on the shelf.

Paused on 16 September: Antenati is returning 403 Forbidden to David's Chrome, the browser that has been reading it all along — the whole site, not one ark, and on the first request rather than after forty-six. That is a block on the connection and not the eight-kilobyte cookie this archive diagnosed in September. Worth retrying later or from another network; nothing else about the reading has changed.

Catalogued and linked; two read. 1806 and 1807 are the years Mione does not have.

Paused on 16 September: Antenati is returning 403 Forbidden to David's Chrome, the browser that has been reading it all along — the whole site, not one ark, and on the first request rather than after forty-six. That is a block on the connection and not the eight-kilobyte cookie this archive diagnosed in September. Worth retrying later or from another network; nothing else about the reading has changed.

Mieli turned up by accident — its arks sit immediately before Mione's, which is how the one-comune-per-block assumption was caught. The Val di Gorto has a dozen comuni and this archive has catalogued three. A De Franceschi who left Mione did not necessarily leave the valley.

Paused on 16 September: Antenati is returning 403 Forbidden to David's Chrome, the browser that has been reading it all along — the whole site, not one ark, and on the first request rather than after forty-six. That is a block on the connection and not the eight-kilobyte cookie this archive diagnosed in September. Worth retrying later or from another network; nothing else about the reading has changed.

27 September 2026 — the 403 is gone, and the comuni have been surveyed.

The block was client filtering, not the network. Antenati still returns 403 to a plain curl and 200 to a real browser, on the same machine and the same connection. It was read here from the in-app browser — not by putting a browser's User-Agent on a script, which would be dressing up a client the site is deliberately turning away.

The name index does reach Carnia. cognome=De Franceschi returns 418 rows across Italy, and the Località facet carries Carnic comuni.

Of twenty Carnia comuni tested, three hold the name: Paluzza, Mione, Ovaro.

Five more are proven to be indexed and to hold none of it — Comeglians, Prato Carnico, Rigolato, Tolmezzo, Cercivento. That distinction cost two controls and is the whole value of the survey: a nil from this portal means «no De Franceschi» only if the comune is in the index at all, and «Rossi» returns nothing anywhere in Carnia, so it proves nothing. «Della Pietra», «Puntel» and «Craighero» are Carnic and they do.

Eleven remain unknown — Forni Avoltri, Sutrio, Arta Terme, Ampezzo, Socchieve, Enemonzo, Raveo, Villa Santina, Zuglio, Treppo Carnico, Paularo — because none of the three control surnames matched there either. They are not negatives.

And Paluzza's twelve rows are already in this archive, every one. Francesco of Naunina 1807, Maria Anna 1810, Pietro Antonio 1810, Cattarina 1811, Daniele 1782, and the mothers Domenica Englaro, Maria Delli Zolti and Maria Silverio — all carried under src: record, which means these registers were read here before the portal was asked. The survey confirms the existing reading against an independent index and adds nothing to it. The remaining value of this item is Mione, Ovaro, and the eleven comuni whose coverage is still unknown.

One measurement was thrown away getting here: every comune reported «10 results» because the count was being read off «10 risultati per pagina». The real totals are in the facets.

The survey is complete, and the method that completed it is simpler than the one that started it.

Guessing control surnames — «Rossi», then «Della Pietra», «Puntel», «Craighero» — answered five comuni and left eleven unknown, because a nil proves nothing when the control surname may simply not live there. Querying a comune with NO surname at all asks the index directly whether it holds that place, and answers every one.

Eighteen Carnia comuni, settled:

• The name is present in TWO — Paluzza and Ovaro. (Mione is a frazione of Ovaro and returns Ovaro's rows.)
• Thirteen are indexed and hold none of it: Comeglians, Forni Avoltri, Sutrio, Arta Terme, Socchieve, Raveo, Villa Santina, Zuglio, Paularo, Prato Carnico, Rigolato, Tolmezzo, Cercivento. These are real negatives.
• Three are not in the name index at all and remain unknown: Ampezzo, Enemonzo, Treppo Carnico.

⚠ What this does NOT say. The Napoleonic civil registers run 1806–1815. This is a map of where the name stood in those nine years — not in 1742, when a Svetvinčenat clerk first wrote «ex Cargn.» against a Defranceschi. Two comuni in 1810 is a lead towards the Carnic homeland and not a finding about it, and the parish registers that would speak to the 1740s are a different body of records entirely.

Done, and the rule changed under it. All eleven living people the archive names are placed — eight at Johannesburg, five at Brisbane, two in both — and eight more came in from the tree on 15–16 September: Lidija, Peter, Renata and Ursula of the Crikvenica and German households, and the Australian branch. Twenty-six living people, and the build gate confirms zero dates reach any page.

Struck the day it was raised. The question was whether Erman is a by-name on a De Franceschi house. It is not: his 1809 marriage act, already in this archive, gives him as son of the late Pietro Erman and Maddalena née de Franceschi. The compound is his two parents. Logged.

The marriage index calls Osvaldo's wife Valentina Prencis on 6 August 1807. The birth act calls her Cattarina nata Crencis on 15 October 1810. Both readings are held open. The Mione «Morti» would settle it if the first one died.

Paused on 16 September: Antenati is returning 403 Forbidden to David's Chrome, the browser that has been reading it all along — the whole site, not one ark, and on the first request rather than after forty-six. That is a block on the connection and not the eight-kilobyte cookie this archive diagnosed in September. Worth retrying later or from another network; nothing else about the reading has changed.

27 September 2026 — half of it is settled, and the half that is settled is the surname.

Antenati's name index gives Osualdo two children by the same wife, five years apart: Gio Batta, born Ovaro 15 October 1810, madre Cattarina Prencis; and Maria Maddalena, born Ovaro 4 November 1815, madre Catterina Prencis.

So «Crencis» is gone. The archive had «Cattarina nata Crencis» off one hand on one act. Two acts, two clerks, five years apart, both write Prencis — which is also the surname the 1807 marriage index gives Valentina. The wives of 1807 and 1810 share a surname; the question was never about that.

Valentina or Cattarina is still open, and the new act does not close it. Cattarina is his wife in 1810 and still in 1815, which fits a second marriage as comfortably as it fits one woman recorded under two names. The Mione «Morti» would still settle it — item 4 — and that is now reachable again, because the 403 is gone.

And the name index cannot finish it — that is now proved rather than assumed.

Ovaro's Antenati index is BIRTHS ONLY. Four surnames tested: De Franceschi 10 birth acts, Prencis 14, Gortani 1 — and zero marriage acts and zero death acts between them. So the 1807 marriage that names Valentina is not in this index at all, and neither is any death.

Nor can the women's own births be there. The Napoleonic registers at Ovaro run 1806–1815. A woman marrying in 1807 was born in the 1780s. The fourteen Prencis births are all children of 1810–1815 — Biaggio, Domenica Anna, Anna Lucia Luigia, Giovanni Battista — not the generation in question.

The route is unchanged and it is item 4, the Mione «Morti» read as images. The index has given all it holds.

He witnesses Lucia Santa's wedding in 1815 and the act calls all four witnesses «non parenti, né affini». Her father was Osvaldo Antonio de Caneva; Daniele, forty-nine, was born about 1766. Either two Osvaldos at Liariis, or the formula is boilerplate.

Paused on 16 September: Antenati is returning 403 Forbidden to David's Chrome, the browser that has been reading it all along — the whole site, not one ark, and on the first request rather than after forty-six. That is a block on the connection and not the eight-kilobyte cookie this archive diagnosed in September. Worth retrying later or from another network; nothing else about the reading has changed.

The single most valuable letter this archive has ever drafted. Mione's civil births stop at 1810 and there is no civil registration at all before 1806 — so Paolo, Osvaldo, Francesco, Matteo, Felice Santo and now Pietro were all baptised in a book only Gorto holds. The draft is written, in Italian, and ready to send.

⚠ 29 September 2026 — David: «not sending letters yet.» This item waits on that decision and is not to be raised again until he lifts it. Recorded once, here, so no session re-opens it by suggesting the letter afresh.

DAPA has answered twice — for Svetvinčenat and Pićan — and both times asked for the same completed form: applicant's details, delivery address, billing details including an OIB. A third sits with DAZD. This archive does not hold that data and will not invent it.

⚠ 29 September 2026 — David: «not sending letters yet.» This item waits on that decision and is not to be raised again until he lifts it. Recorded once, here, so no session re-opens it by suggesting the letter afresh.

L-M27 rests on a father and a son, which is the same opinion asked twice. One cousin reading his paternal haplogroup off an account he already has settles it either way. The message is drafted; the Gračišće reunion families are the first ask.

Done, and it was done before this row was raised. The undigitised sets the forty-nine out place by place and holder by holder with the DAPA shelf numbers, and carries a draft request in Croatian with the blanks named — the DAPA number and the years are what the office works from. out/dapa-zahtjevi.md carries the filled-in guidance for the two forms.

What is left is not writing but sending, and that waits on David's own details, like every other DAPA form.

Done: site/src/data/reads.json. A row names a place, a volume in the words the panel prints, a state — read or part — and what was read. The builder stamps it onto the volume and reports any row that matches no volume rather than dropping it, which is how the Vodnjan title was caught wrong on the first run. Four volumes pinned so far; six places now have a book of their own opened.

Found, walked and catalogued on 15 September. The collection's own waypoint list names it — ten confession roots, and «Evangelical (Evangelička crkva)» is sd_4, waypoint 9R27-7M4:391644401. Six parishes, seven volumes, all now catalogued with their arks harvested and openable from the map. Brekinjska and Hrastovac stop being grey, and every grey marker left on the map is a checked negative rather than an admission. On the search register.

Done, and it widened by a different route than expected. The note said better name matching would turn one finding into forty. It did not need to: the providers multiplied instead, and the test now stands at 27 places against six — Istria against Pazin, Međimurje against the Slovenian collection, Croatian Istria against the Italian one, Delnice against the main Croatian shelf. 294 volumes before 1890 listed by one holder and unfilmed by the other; 393 after 1890, which is privacy and not a hole.

And it was wrong in the other direction first. The library catalogue had crept into the comparison, which inflated it to 78 places and 1,572 volumes — a bibliographic line spanning three centuries being read as a claim about a particular year. fs-cat is now excluded from the test, because a catalogue is not a shelf.

Walked, and there are four congregations on this archive's ground, not two. Pola — Taufbuch 1873–1913; Görz — Taufbuch 1860–1913; Triest AB — four registers 1848–1918; Triest HB — three, 1900–1918. Nine volumes, free and online.

Pola is the one that matters and the row had not heard of it: this family's southern-Istrian cluster is Pula, Ližnjan, Fažana and Medulin, and Pola in 1873 was the imperial naval base filling with incomers.

The family is Roman Catholic and an Evangelical baptism book is unlikely to hold it — but unlikely is not checked, and this map already carries six Evangelical parishes walked on exactly that argument. On the search register.

./scripts/publish.sh builds and audits clean — 1,231 pages, every link and anchor — then fails on MISSING OUTPUT: /404/ and refuses to push. An untracked 404.astro builds to 404.html rather than 404/index.html. It belongs to the session that owns the cross-site UI. Everything here is being committed and pushed by hand around it.

Fixed, 20 September 2026 — by this session, in this archive's own script. publish.sh derived each page's expected output as dist/<slug>/index.html, and Astro emits 404.astro as dist/404.html by design. The check now accepts either spelling:

```
[[ -f "$out" || -f "site/dist/$slug.html" ]] || { echo "MISSING OUTPUT: /$slug/"; missing=1; }
```

It was never the other session's to fix — the 404 page was correct all along and the checker was wrong. Publishing by hand around it has stopped; every commit since has gone through publish.sh with all seven gates.

425 volumes in the Istrian and Kvarner parishes this family is documented in, each now opening at page one from the research map.

Struck: it answered a question nobody was asking. Shading by how many volumes sit on a shelf looked better than the saturated three-state scale it replaced, and said less — a parish with forty volumes nobody has opened and a parish with two nobody has opened are in the same state. The colour now says how far we have got: never opened, started, read right through. Depth moved into the panel, in words, where it can sit next to «nobody has read any of it» without pretending to be it.

Seventy-four volumes for Mione, Ovaro and Mieli, every one verified by fetching its own catalogue page. Its arks sit on a different host from everything else, so the map sends a reader to the archive's own list rather than dressing them as FamilySearch links.

Matricula holds no Croatian parish at all. Twelve countries in its fonds index and Croatia is not one of them; its place search returns nothing for Senj, Karlobag, Rijeka or Fiume. The line was written from memory and never checked. Logged as a correction.

Six places are described by more than one provider, and they are the only places a gap can be seen. 147 volumes are listed at Pazin, from before 1890, that FamilySearch has not filmed — and 313 more after 1890 that are the privacy cut-off and not a gap at all.

The best single hour of the day. Felice Santo's baptismal extract of 2 November 1792 names «Pauli filii Petri De Franceschi mionensis» — a grandfather this archive did not have — copied out of the Pieve di Gorto's own register and certified at Luincis.

Both. And the gap turned out to be narrower than published: Ovaro kept 1807, so it is one lost Mione volume rather than a hole in the system.

Seventeen — from Felice Santo the weaver and Lucia Santa de Canova down to the two Luint tradesmen who witnessed a birth in 1810, and the priests who certified the extracts.

Six spellings, three women, one Friulian house-name — «in cjasa da Prences». And a 1792 priest's Latin writes it Prencis, with a P, which is the earliest and most careful hand of the lot.

The hundred and fifty-four — by place, by holder, with the DAPA shelf numbers and a draft in Croatian. 34 of them are in parishes on our own line.

The first Antenati catalogue named fifty-nine volumes from contiguous ark ranges, assuming the portal numbers one comune at a time. It does not — an_ua667510 to 667513 are Mieli, a different comune, sitting immediately before Mione. Every volume is now fetched individually or it is not claimed. Logged as a correction.

A marker used to list its volumes in whatever order the providers were walked. It now shows a coverage timeline — Births, Marriages, Deaths — with every hole named and measured. 277 real gaps across 160 places; 1,002 places have none.

A one- or two-year hole is called a labelling artefact rather than a loss, except where every volume covers a single year — the Napoleonic comuni — where a missing year is a missing book. That exception is what makes Mione's lost 1807 show as three gaps and not three shrugs.

The «also written» line used to carry nothing but the catalogue's own noise — «Fažana, Croatia». It now carries what the place was called when the register was written: Rijeka as Fiume and Sankt Veit am Flaum, Pazin as Pisino, Gračišće as Gallignana, Svetvinčenat as Sanvincenti. 609 of the 1,190 mapped places have one.

Taken from the GeoNames alternate-names column already held offline, with the airport codes, the Mandarin and Cyrillic romanisations, and the modern administrative wrappers («Grad Vodnjan») filtered out — and with spellings that are the same name without hačeks (Fazhana, Svetvinchenat) folded away rather than listed as identities.

Checked against the full 1,333-parish walk of Croatia Church Books 2040054: Gračišće, Gologorica and Pićan are not in it under any spelling — nor Lindar, Beram, Trviž, Cerovlje, Zarečje, Grdoselo or Kašćerga. Boljun, Kršan and Kozljak are. It is a patterned hole across the whole Pazin hinterland, which is the ground this family comes off.

Closed as answered rather than solved. All 3,488 FamilySearch collections have since been walked and none of them holds these parishes — the hole is real and it is FamilySearch's, not a spelling problem. Gračišće, Gologorica and Pićan are at Pazin and nowhere else, which is why their 32, 22 and 4 volumes are catalogued and unopenable, and why they lead the undigitised. The next move is the DAPA request form, which is blocked on David's own data, not on more searching.

Green on what is done, red on the gaps, amber on what is part-done — David's correction, applied to the panel as well as the marker.

A record-type heading turns green when every volume of that kind has been read, amber when some have, and stays plain when none have. Individual volumes go green when they have been read, in the collapsed list and in the clickable «open at page one» list alike. Gaps keep their red.

Mione now reads at a glance: BIRTHS · ALL READ in green, MARRIAGES · 6 of 18 read in amber, and 1807 in red three times over. All of it done from outside the shared Atlas component, which this session does not own.

Meljani has a single register and the panel was showing it four times: as a record-type heading, as a coverage span, as a volume row, and again in the kit's own «open at page one» list.

It is now one line, with the link moved into the row it belongs to and the leftover list removed. The span row only appears where there is a gap to show; the fold only appears where there are more than four volumes. Done from outside the shared component, by matching the two lists on a label the builder deliberately makes identical.

Walked. Six reels, ~3,485 images, and no parish structure at all — the collection browses by film number only. They hang on Delnice itself rather than pretending to a precision the collection does not have. Which parishes each reel covers is a question for the catalogue record, not the browse tree.

Struck: it does not cover the Littoral. The collection has no browse tree — it is a name index of 6.85 million records — so it was searched instead. 289 Defranceschi inside it, every named place Tyrolean: Bozen, Salorno, St. Florian. Three Istrian surnames tested beside it fall the same way.

Gračišće, Gologorica and Pićan are not here, and this was the best remaining candidate. The negative is recorded in full.

Struck: Croatia-Slavonia never had Hungarian civil registration. The collection browses by 33 Hungarian counties and not one Croatian one, and there is no Fiume in it.

Hungary introduced civil registration in 1895; Croatia-Slavonia kept the parish registers as the legal civil record until 1946. So there is no Croatian civil registration to find — it was never created. The negative, and why it explains Međimurje.

Done. 58 places, 409 volumes, 32 of them Croatian, across three collections filed by Italy, Slovenia and Croatia-by-film.

An exonym table puts each on the same dot as its modern name — Csáktornya→Čakovec, Perlak→Prelog, Buie→Buje — with the catalogue's spelling kept in «also written», because that is the name you must search for.

It found a gap on day one: Čakovec is missing 1891–1894 in all three record types, the interval between the parish books closing and Hungarian civil registration opening. Neither catalogue could show that alone.

Done, and it is exhaustive. All 3,488 collections listed, every title read, and 44 opened — every Croatian, Austrian, Hungarian and Slovenian one, plus the Italian collections for Gorizia, Trieste, Udine, Venezia, Belluno, Treviso, Padova, Verona and Vicenza.

Five hold places on our ground and all five are now on the map. Thirty-nine hold nothing, and each is recorded as a negative so nobody opens it twice.

There is no Serbian, Bosnian, Montenegrin or Yugoslav collection at all, and Croatia has exactly two.

It had a page already — Trentino — and what it was missing was the registers. That page opened by saying every one of these people arrived through the family tree and no register had ever been opened for any of them. The Austrian collection walk had found 147 indexed entries, 1608 to 1915, and nobody had put the two together.

146 of the 147 are baptisms. 132 name both parents and 126 name a spouse — more filiation than this archive holds for most of its own Istrian century. They are on the page now.

And they sharpen the argument rather than softening it. Nineteen in Carinthia and sixteen in South Tyrol is a scatter, not a cluster, and the surnames run Franceschini and Franzeschini — Italian names that reached Austria by a different road from the Istrian one. A hundred and seven of the 147 carry no town at all, which is the same index failure this archive already solved for Croatia by resolving the image ark to its film; that resolution has not been run here.

Two collection walks in one evening produced the same structural fact from opposite ends: Hungary registered civilly from 1895 and Croatia-Slavonia did not, keeping the parish books as the legal record until 1946.

Written and published on the method page, where it belongs: the Nagodba of 1868 left Croatia-Slavonia its own administration, so Hungary's 1895 civil registration did not reach it and the parish books stayed the legal civil record until 1946. It explains why FamilySearch's Croatian collection is church books, why Pazin's list is parish registers, and why Međimurje is the exception — it was governed as part of Zala county, which is Hungary proper.

FamilySearch has 257 Italian collections; the walk opened the nine that touch the Adriatic and the Veneto. The rest are Naples, Sicily, Piedmont, Puglia — irrelevant to this family's ground, and squarely the business of the Falco and Mazza archives.

Decided against, and the reason is scope rather than effort. Naples, Sicily, Piedmont and Puglia are not this family's ground in any century the archive can reach, and walking them would fill this map with a quarter of Italy to no purpose. The nine Adriatic and Veneto collections that could hold Istrian or Carnian material have been walked and are on the collections walk.

The list is worth handing over rather than working: the Falco and Mazza archives sit on southern Italian ground and the same walk would be their whole first week. Recorded here as a decision so that the next session does not re-open it.

A second layer on the research map that is not about books at all: 160 towns, 956 placements of documented people, 106 of the towns outside Croatia — Torrington, Memphis, Pfullingen, Broken Hill, Brisbane. Lifted whole out of atlas.json rather than geocoded a second time, stripped of its film lists so no register is offered twice, and switched off by default with a violet filter button that turns it on. 43 towns carry both a people marker and a shelf marker.

The research map, the atlas and the graves map were three maps of the same ground with three geocoders and three legends. 1,235 + 160 + 64 markers are 1,358 places; 84 answer more than one of the three questions and 12 answer all three. The colour switches between the questions; the panel never does. /research-map/, /atlas/ and /graves-map/ redirect.

The living rule changed on 15 September: name and town, never a date. All eleven living people the archive names are now placed — David and Cheryl at Johannesburg and Brisbane; Cristina, Alessio and Mateo at Brisbane; Ada, Horst, Mark, Gayle, Nicole and Kirsten at Johannesburg, on David's word of the same day. Robbie — deceased, no date held — is placed at Johannesburg too, marked inferred: his wife and both his daughters are there and nothing in the archive names a town for him.

Folded into one row. He was Mario De Franceschi 1938–2013 off a Find a Grave memorial and Mario Defranceschi 1938–2013 out of MyHeritage 4507883 — two slugs under /who/, two dots in Brisbane, one man. The surviving row carries both spellings and both sources.

The heading on this row was wrong twice over. It said «David's father», which is what the mis-flagged rows had made this archive believe; his father is Ivan Anton Defranceski, 1951–2026. Correction 79.

Answered: yes. He is Anton Mario De Franceski, born 21 August 1944 at Reutlingen — Zdravko and Ruža's only child, four years old on the General M B Stewart. Eight NAA items say so, including A11534, 58–60, which carries all three dates of birth in one line, and D4881, his own arrival card, which gives the middle name and the Australian spelling. Two corrections came out of it: he was four on the ship and not nine, and «Zorafko» at Horsham is Zdravko, so that is one burial and not two.

The card runs 13 April 1949 to 31 December 1976 and is held at Adelaide; it should carry addresses, and it is the cheapest route to how a Horsham boy ended up at Ipswich. Ryerson Index — free, covers Australian death notices for 1999 — was returning «Server Overloaded» on 15 September and needs another run; a funeral notice would name a wife and children, and would settle the two-year gap between the cemetery's age and four documents.

The roster's family-file source is a GEDCOM of July 2018 — 8,286 people. The live tree has 15,643. A note of 7 September already said the export was retired and read 459 individuals of the surname out of the live tree; that reading was never imported. The Australian branch went in on 15 September — fifteen people — because David asked why his cousin's sister at Horsham was missing. The rest of the tree is still outstanding, and every one of the 513 roster rows carrying a 2018 identifier needs checking against it.

16 September: the live tree is finally readable. The reconciliation is published and names the 668 people the roster does not have and the 14 dates the two sides disagree on. Importing them is the next pass, and it is a judgement call each time — a tree absorbs other people's Smart Matches, and this archive does not publish somebody else's guess as a fact.

20–21 September: the import happened, and the number is 84, not 668.

Eighty-four people came out of the live tree and into the roster — the ones the reconciliation found in the tree and in no pen of David's own. The other 486 of the 668 were already inside his pens, tested and parked, so importing them would have duplicated work rather than added it. Item 93 is done.

Every imported row carries its own caveat in its own note — «Tree-derived: no register has been read for this person here», graded inf, with the tree named as the source and the date of the reading. That is the whole point of the pass: the archive now holds these people and holds the fact that nothing has been proved about them.

And one of them stopped being tree-derived the next day. Dionigi de Franceschi, b. 1927, Gračišće came in on 20 September with that caveat; on 21 September the Dachau entry register named him — «DEFRANCESCHI, DIONIGI, born 17 September 1927 at GALLIGNANA» — and the caveat came off. That is what the caveat is for: it marks the rows worth attacking.

What remains on this item is the other half: the 513 roster rows carrying a 2018 GEDCOM identifier, each of which needs checking against the live tree. None of that has been done.

Born at Crikvenica in 1915, married name Oreščanin, died at Belgrade in 1944. That is the whole of it. A Croatian woman of twenty-nine dying in occupied Belgrade in 1944 is a question, not a fact — and this archive has never asked it. Arolsen, the Belgrade city registry and the Yugoslav victims' census are the places to start.

The tree adds two husbands and deepens the question. Josip Geiger, 1911–1936, dead at twenty-five when she was twenty-one; then Branko Oreščanin, whose name she carries. And her mother Ursula died at Belgrade too, in 1975 — two women of one family dead in that city, thirty-one years apart. There is now a second surname to search Arolsen and the Yugoslav victims' census on.

Arolsen searched on 16 September and she is not in it. No person record for Defranceski Anka; no match under Oreščanin or Geiger either. Her brothers are all over that archive — the surname returns 79 person records — and she is not, which says she died at Belgrade outside the German system. The Belgrade city registry and the Yugoslav victims' census are what is left.

And the 79 is a finding on its own: this archive has read exactly one Arolsen file.

20 September: a second door closed, and it narrows the question rather than answering it. The Jasenovac named victims list was searched on the surname in three spellings — all zero — and on both her married names. Oreščanin returns four people, none of them a woman born at Crikvenica in 1915; Geiger returns two, neither her first husband, who died in 1936 in any case. A control on a common surname was run first, so the zeros mean something. On the research log.

Two systems have now said no. Arolsen: she is outside the German records. Jasenovac: she is outside the Ustaša camp records. That leaves the city itself, and it points the search at Serbian holdings this archive has never opened:

• the Belgrade city death register for 1944, at the Matična služba;
• the Historical Archives of Belgrade, which hold the Banjica camp inmate books — the one Belgrade institution that would record a detained woman;
• the 1964 Yugoslav census of war victims, at the Archives of Yugoslavia, which counts civilian deaths by cause and place;
• and, because it is the single likeliest civilian cause in that city and year, the casualty lists of the Allied bombing of Belgrade, heaviest on 16 April 1944, Easter Sunday, when something over a thousand civilians were killed. inf — that is a hypothesis about a date, not a claim about her; nothing in this archive yet ties her to it.

⚠ 29 September 2026 — one of the two remaining routes is now closed, with a negative.

The [Museum of Genocide Victims' census at Belgrade](https://baza.muzejgenocida.rs/) — 659,122 records — does not hold her. Seven spellings tried and named in Searched; none returns an Anka. She is not in the Yugoslav victims' census.

That leaves the Belgrade city registry, which is the one route this item has never had a way into. It needs a written request, and it is David's to send.

What the same search did give: three others of the name, none of them her — Anton of Gologorica who died at Dachau, and two young men of Vodnjan.

Three living children of the Australian branch are now named on the map with no dates and a town — Charlotte Rita Baviello, Louis Hendy and Lewis Rose, alongside Alessio and Mateo. That follows the rule as David set it: name and town, never a date. They are other people's children rather than David's, and if he would rather they were not named at all, it is one line to take them off.

David exported it on 16 September, and the one click did what a fortnight of crawling could not. 15,643 people, 4,777 families, 1,413 of the surname — not 798, and not 459. The walk's count was an undercount too, because following family links only reaches people the tree has joined up.

The whole reconciliation is published: 415 agree, 14 disagree, 668 are new here and 780 are ours alone. The GEDCOM itself is not committed and will not be — it carries dates of birth for 117 living people of the name, and this site is public.

689 of the 1,413 carry a citation, not 313. They are not where the crawl was looking: MyHeritage files its own Smart-Match plumbing under «SOUR», and the citations David typed by hand sit in «NOTE» records. Reading the wrong tag found 272. Reading the right one found 689.

Between them they name 405 distinct FamilySearch arks — 311 indexed records and 94 film images — and 328 of those arks appear nowhere in this repository. What they cite, and which of it is new.

One typo, one person. Defraneski returns Lisa and nothing else; De Francesci, Defrancheski, Defranceshi, Defrancesky, Defranchesci, Defrancesci and Franzeschini all return zero. The blind spot was real and it was one letter wide.

Sideways find: Franceschini returns a Brazilian-forename group — Aurelio 1892, Turno 1894, Aristeu 1901, Tosca 1907 — which belongs with Trentino.

Read, and it is better than the row predicted. David built a node called «PLACEHOLDERS Defranceschi's for Investigation (not real)» with twenty-three children, one per country, and under the Croatian one seventeen more, one per town — including Gologorica and Gračišće. Australia is filed by state, and Mario of Brisbane hangs under QLD: David had filed him as unconnected himself, years before this archive claimed him and had to take it back. Published on the diaspora page.

Run, and twelve rows failed it. The test is written down for the first time: «mine» means this archive can NAME the path from this person to David — a place on the direct line, a named relationship, or a household the archive has drawn. A shared surname on a gravestone is not a path. Twelve Find a Grave memorials came off the flag — Millbrook, the Bronx, East Farmingdale, Middle Village, Hartsdale, and the Lilydale, Fawkner and Parkville Italians the diaspora page had already filed under a different migration. 187 rows keep it. Correction 80.

Done. About now carries the live tree as a source in its own right — 15,643 people, tree 4, read 7 and 15 September — and the July 2018 export is marked SUPERSEDED in its own title, with the reason on it.

Done: scripts/gazcheck.py. Every map builder now dies with a message naming the path it was given when its gazetteer comes back short. Tested with a wrong directory: exit 2, nothing written.

Done, in the same file. A builder that places fewer places than its own last output refuses to write and says by how many. --allow-shrink is there for a fall that is real and intended.

Named, and there are three children rather than two. Ursula (Korn), born at Urach 1940; Renata (Buck), born at Metzingen 1945; and Peter, whom the Arolsen cards do not mention at all. Their mother is Luise Salzer, born 1919. All three are living — eighty-six, eighty-one, and undated — and are on the archive now by name and town, which the present rule allows and the old one did not.

Asked, and the answer is that there is no answer in the tree. She is a spouse node with no dates, no places, no children and no maiden name; «Koka» is a pet name. She is on the roster now as exactly that, and as an open question — David's grandfather had twenty-six years at Senj after the divorce and this is the only trace of them.

Codron is a later husband, not a maiden name. Hedviga married Santou (Shemtov) Isaac Codron, 1913–1997, and the tree names a third husband, Ladislav Carr; her parents are Ljubomir Blažević and Milka Papić.

And the question paid for itself twice over. She reached South Africa in April 1975, from Zimbabwe — a whole leg of this family's journey that no page of this archive had — and her birth date is settled at 3 August 1926, where the archive had carried two transposed dates and preferred neither.

Naturalisation at Melbourne in 1978, twenty-nine years after they landed and a year before he died. The date and the address no longer depend on the file: the Commonwealth Gazette prints them — 21 July 1978, 19 Anderson Street, Horsham, read on Trove. What is still only in the file is occupation, arrival particulars and a photograph, and this archive has no photograph of either of them. Access status «not yet examined», which means it has to be asked for before it can be read.

Checked, and the answer is none of them. All eight items are «Open» and not digitised — A11534, both B44 naturalisations, the three A2571 alien registrations and the two B78 cards. «Open» in RecordSearch means readable, not scanned, which is exactly the distinction this check existed to draw.

It turns the whole Australian side of this family from a search into an order, and the two B44 files are still «not yet examined», so access has to be applied for before a copy can be. They should go in one request rather than eight.

Turned into an errand rather than a search, because it was never a search. The free Victorian index returns a name, a year and a registration number; the parents are on the certificate, which has to be applied and paid for. Zdravko's 1979 death is inside the open window and is written up as an errand; Ruža's 1996 death is not — Victoria releases deaths at thirty years.

Separately: the BDM search form could not be driven from either browser here. Its Angular model ignores values set programmatically, and real typing needs a browser pane this session does not have. One minute of David's own browser.

Done, 17 September 2026, and it paid twice. Commonwealth of Australia Gazette. Periodic, No. P2 of 16 February 1979, page 66, under VICTORIA: «Defranceski, Zdravko, 19 Anderson St, Horsham 3400, 21 7 78» and Roza on the line above — naturalised 21 July 1978 at Horsham, not Melbourne. And No. P8 of 25 October 1978, page 303, under SOUTH AUSTRALIA: «De Franceski, Anton Mario, 7 Calomba Cres, Happy Valley 5159, 15 2 78» — the son, five months earlier, in another state. The whole story.

Run, and the answer is nothing — weakly. No result for the surname in Trove's newspapers across the 1940s or the 1950s. Trove's newspaper digitisation mostly stops in 1954 and whether the Wimmera Mail-Times is held for 1949–54 has not been established, so this rules out the easy version and not much more. On the search register.

Both are in the live tree, both deceased, and neither is in this archive. Ruža was born at Fužine on 13 February 1922 — her own town certificate says so against a single Arolsen entry that reads 2 May — and a second place, Belnica, appears on the Neuhausen list. The Hreljac line enters this family through her and has never been opened.

Checked 15 September: Fužine's films stop at 1889 and Ruža was born in 1922, so her own baptism is not on any film this archive can reach. Her parents Lovre and Dragica married before that, though, and Fužine marriages 1778–1857 and births 1725–1889 are clickable from the map — that is where to look for the Hreljac household, not for her.

17 September: Arolsen gives her a date and a spelling. «DEFRANCESKI · Rose · HRELJEC · Fuzinä / Belnica · 02/13/1922» — born 13 February 1922 at Fužine, with Belnica beside it. The German clerk heard Hreljec; this archive has Hreljac. The Fužine registers settle which, and now they can be searched on a date.

20 Sept — searched, and the item points at the wrong parish. No Hreljac record at Fužine at all. The surname clusters at Grižane, Tribalj and Bakar — the Vinodol, the same valley as Crikvenica and Klenovica.

One candidate: «Lovra Hreljac, born and baptised 4 September 1883 at Grižane», son of Stepan Hreljac and Agna Barac (6YBK-YPMM). Ruža was born 1922, so the age fits exactly. inf — nothing ties him to Fužine or to a Dragica Vlahović, and no such marriage is indexed.

Read Grižane, not Fužine. Fužine may hold the wedding; the household is Vinodol.

21 September 2026 — FamilySearch is ruled out, and the ruling-out was tested. Exact-surname Hreljac across Croatia born 1918–1926 returns 0; the same query without the date range returns 340, which proves the service was answering rather than quietly failing. A place query on Fužine returns nine records and not one of them is a Fužine record — the word appears nowhere on the page. The indexed Hreljac material is Grižane, Lussinpiccolo and emigration sets. Fužine's registers are not in this index at all.

This item should merge into the Rijeka enquiry rather than stand alone. Item 141 already asks Rijeka for the Klenovica and Ledenice books; Fužine is the same archive and the same period, and Ruža's parents are now a register question, not a search question.

Both answered. Fužine is on the map with five volumes, all of them now clickable — births 1725–1778, 1778–1857 and 1858–1889, marriages 1778–1857, deaths to 1916 — and it is in the main Croatian collection, not in Delnice. Belnica is on no catalogue this archive holds and is not on the map; it is a hamlet rather than a parish, and whatever it means on Ruža's Neuhausen entry it is not a register office.

Answered, and it is an order rather than a search. FamilySearch's Crikvenica films are cut at Rođeni 1896 — the hundred-year line as it stood when they were made — so all five children of Anton Rudolf and Ursula were born after the end of the only collection this archive can reach for free.

They are all on one reel. HR-HDA-1448, R 1909–1926, roll ZM-34J/362, mark M-6957, Matični ured Crikvenica, digitised: Josip Petar 1911, Petar Vid 1913, Anka Nera 1915, Zdravko 1921 and Ivan Anton 1925. Written up as an errand with the Status animarum and the 1909–1947 marriages, which come off the same collection and should go in one request.

Crikvenica births 1897–1908 are in neither collection — that stretch holds Ruža Jarmacki born 1903 and the POLD… of the monument, and it needs the parish office or the Rijeka archdiocesan archive.

The archive and the tree both say «circa 1943, Crikvenica». His son Zdravko was taken by the Partisans that same year. A death in occupied Croatia in 1943 with no date is a gap worth closing, and the Crikvenica death register is the place it closes.

Checked: it cannot be searched, only asked for. Crikvenica deaths on film stop at 1908; the 1943 register sits with the Matični ured, and the errand for it has been standing since before this session. The same office holds the 1909–1926 births.

17 September: an Anton born 20 January 1890 is in Arolsen's 2.3.3.1, which is the ITS collection of documents whose names were never captured into the main index. That is Anton Rudolf's own birth date. Why a man who died at Crikvenica in 1943 has a record in a tracing collection is worth one look — it may be a search his own children filed for him after the war, and a tracing enquiry usually states the death it was answering.

Josip Defranceschi, generation 4 — Gračišće to Crikvenica. The archive draws the move and has never explained it. It is the second unexplained jump in the line, after Mione to Gologorica, and unlike the first one it happens inside the nineteenth century with registers on both ends.

21 September 2026 — out of reach of full-text search, measured. FamilySearch full-text indexes Italian civil registration, not the Croatia Church Books: 97 collections sampled from its own results carry zero Croatian entries. This item is a Croatian church register and no query will reach it. It stays an archive question.

Anton Rudolf is «born 20 January 1890, Ledenice / Klenovica 22». A house number is an address the land registers and the church's status animarum can both be searched on, and this archive has done neither. It is the Kvarner equivalent of Gologorica's house 17.

20 September 2026 — the house is identified, from the Ledenice baptisms.

The register names one house four ways between 1861 and 1877 — Klenovica, Rljinovica, Rljenovice, Povile — and the number is 22 every time, with the same parents, Josip Kalanj and Rosalija née Perković. The 1861 entry spells it out: «Ledenice in Klenovica N° 22».

Anna Kalanj was born there on 3 January 1864, and her son Anton Rudolf was born there on 20 January 1890. The 1890 remark — «kod roditelja žene Ane» — is naming this household, and the godfather Jakov Kalanj of house 22 belongs to it.

Still open: whether Povile 22 is literally the same building or a neighbouring hamlet's house of the same number. The parish, the number and the parents are firm; the building's own village name is not settled by these entries alone.

21 September 2026 — out of reach of full-text search, measured. FamilySearch full-text indexes Italian civil registration, not the Croatia Church Books: 97 collections sampled from its own results carry zero Croatian entries. This item is a Croatian church register and no query will reach it. It stays an archive question.

Overtaken. The harvest finished the whole queue before a priority order was worth building — all 5,482 volumes are linkable, Senj's fifteen among them. Worth doing if the queue is ever regenerated at that size again; not worth doing now.

Done — four volumes pinned, and six places now have a book of their own opened instead of two. Gračišće's baptism book 1667–1745 and Donji Proložac's deaths 1837–1859 are read right through; Svetvinčenat's 1663–1708 and Vodnjan's 1815–1884 reel are read in part. «Part read» now counts as started.

Done. A teal marker prints the rows themselves under a heading that says not one of them is a volume from the shelf above. Žminj shows «Libri žminjski, libar drugi (Žminj 2008) — partial»; a place with more than twelve says how many more.

Done, 17 September 2026, and this row was wrong twice. They could be placed from this repository: data/parish-by-record-2026-09.txt holds the sweep's per-record output — ninety-three parishes, 952 persona ids — and had done since 8 September. All thirty-six placed with no FamilySearch search at all: twenty-five on the id already held, eleven on the twin rule — the same burial indexed twice under ids differing in the last character only.

Svetvinčenat 14 · Pula 8 · Draguč 6 · Rovinj 3 · Vrsar 2 · Šterna 2 · Vižinada 1, four of them parishes new to the register of the dead. A bad pin fell out on the way: Šterna had been resolving to Greece. The correction.

Checked at 375px and it holds — scrollWidth equals clientWidth, so nothing pushes the page sideways. One fix made: the three question buttons size to their own text, which made a ragged left edge on a phone, so below 640px they go full width.

Done by building data/gazetteer.json — 2,456 places distilled from the GeoNames dumps by scripts/build-gazetteer.py and committed, so no builder needs the seventy-seven megabytes of dumps to run. The atlas gained six places by it. The seed keeps only what a gazetteer genuinely cannot resolve, and says so.

Gone. build-atlas.py reads the committed gazetteer; the seventeen hand-checked places the old file carried have been folded in and marked as hand-checked. The file is read if it happens to be there and nothing breaks when it is not.

Hedviga «Seka» Blažević was born at Senj in 1926 and married Ivan Anton there; both registers are with the Matični ured, past the end of every film. The errand is drafted and waits on David's own details, like the DAPA forms. The live tree carries her as «Blažević (Codron)», which implies a marriage this archive has never heard of — the same register would settle that too.

Nine printed titles stand pending on the search register and they are the secondary literature of this family's own historian: Camillo's Storia documentata della Contea di Pisino and Povijest Pazinske knežije, Carlo's Uspomene, Bertoša's two, Jelinčić, Gentileo's bibliography, the 1788 visitation and the Atti e memorie obituary. Four of them are one order from the Pula university library. They are the cheapest unexploited body of material the archive has.

Vergé-Franceschi's «Le cap Corse: généalogies et destins» (2006) is the reference work for Cap Corse families, and the archive holds four Corsican graves at Tomino and Mandolacce. It would settle in one afternoon whether that household is a fourth population of the surname or a branch with a route to it. Not yet obtained.

Listed as «untested» on the search register and left there. FamilySearch's full-text search reads the handwriting of unindexed images, which is precisely the state of every Croatian register this archive has catalogued: 5,489 volumes, none of them name-indexed. If it works on Istrian parish hands it changes the shape of this whole archive. One afternoon to find out.

Blocked on 16 September: the FamilySearch session is signed out. Full-text search redirects to the login page and the persona API stopped answering at the same time. This archive does not enter credentials. One sign-in from David and this runs.

21 September 2026 — it has now been tried, and it worked on the first query. Full-text search runs from a signed-in in-app browser without difficulty: Defranceschi returns 3,967 documents, and the plain query Rotilio Defranceschi produced the Trieste marriage proclamation that settled item 125 — a document no name index holds. This item is closed as answered, and the service should now be run against the rest of this list.

requests/arolsen-td-349822.md — the Arolsen tracing case that stood open on Petar for thirty years, eight documents withheld online. requests/nsk-de-franceschi-archive.md — the De Franceschi archive at the National and University Library. Both are written and waiting on a signature.

⚠ 29 September 2026 — David: «not sending letters yet.» This item waits on that decision and is not to be raised again until he lifts it. Recorded once, here, so no session re-opens it by suggesting the letter afresh.

Udine's Notarile antico B. 2979; the Veritas register for the Aquilone; the «Famiglia Franceschi» notice in Il Rammentatore zaratino 1859; and the Vienna Haus-, Hof- und Staatsarchiv correspondence of December 1918 naming a «Dr Defranceschi». Each is drafted on the errands page and each needs the same thing: a name and an address on the form.

⚠ 29 September 2026 — David: «not sending letters yet.» This item waits on that decision and is not to be raised again until he lifts it. Recorded once, here, so no session re-opens it by suggesting the letter afresh.

Rijeka HR-DARI-382 kut. 1; Zadar's Prijamni ured; Bakar's parish registers; the Stemmi della nobiltà di Buccari; Stare hrvatske novine at dnc.nsk.hr; the Arolsen online collections; and the Museu da Imigração at São Paulo. Seven rows on the search register that say pending and have said it for a week. Several are free and online and simply have not been opened.

Two more of the seven closed on 21 September, and they closed in opposite directions.

Arolsen is open again, and it was worth the wait. On 10 September the collections server timed out; it now answers. Searched under the -schi spelling — which this archive had never tried, having listed 79 records under -ski — it gives 38 people and 9 archival units, and they are a different set. The find is DIONIGI DEFRANCESCHI, born 17 September 1927 at GALLIGNANA — that is Gračišće — prisoner 142393 at Dachau, carried by four separate Dachau series. He was a tree-only name in this roster until today. The full reading is on the search register, with three Istrian displaced-persons households — Dignano, Pola, Trieste — that this archive had never held, and twenty-four new people.

São Paulo is not empty, it is broken. The Museu da Imigração's enrolment search at /acervo/registro-de-matricula returns HTTP 500 for every query — Rossi and Silva included — so a nil there would have been a fabrication. The museum's own e-book says the collection went to the Arquivo Público do Estado de São Paulo in 2011, and APESP sits behind a bot-protection layer this archive does not go around. That row stays pending on purpose.

Still open: four. Rijeka HR-DARI-382 kut. 1, Zadar's Prijamni ured, the Bakar parish registers and the Stemmi della nobiltà di Buccari — all four are requests to an archive or a library, not clicks.

27 September — São Paulo is back, and it is the biggest of the seven. The Hospedaria de Imigrantes enrolment register at museudaimigracao.org.br returned HTTP 500 to every query on 21 September. It now answers, and returns 184 records on the surname stem, each with an arrival date, an age, a relationship and a family number that groups a household across the register. Nothing is harvested yet, and the 184 sweep in Franceschini and other names the stem catches, which are a different surname and have to be separated before a single figure is published.

28 September — São Paulo harvested whole: 184 records, and one of them is the only Austrian in the set.

The Hospedaria de Imigrantes enrolment registers are in data/harvest/saopaulo-hospedaria.json. 184 reported, 184 harvested — nineteen pages, and the file was only written once the two numbers agreed. The pagination parameter is page, and it appears in the markup as &page=, which a [?&]page= pattern will not match.

The stem is not one surname. Of 184 rows, only 73 are exactly FRANCESCHI or DE FRANCESCHI. The rest are FRANCESCHINI (79), Franceschin, Franceschine, Franceschino, Colafranceschi, Gianfranceschi and nine others — different names, and none of them this one.

⚠ FRANCESCA FRANCESCHI, «AUSTRIACA», aged 37, landed 22 May 1888. Of the seventy-three, seventy-two are Italian nationals and she is the only Austrian. Istria, Fiume and Dalmatia were Austrian in 1888, so she is the only person in the whole register who carries the nationality this family would have carried. She is entered as MÃE — mother — under família 01538, so she did not arrive alone, and the family number groups the rest of her household on the same page. The register page is published as a PDF, L011_187.pdf.

Nothing connects her to this family and nothing here claims it. What makes her worth a line is the process of elimination: if any of the 184 belongs to this archive's people, it is the Austrian one.

And the família number does what this archive hoped it would. Four men land together on 27 April 1887 under consecutive numbers — Gio Batta (46700), Giuseppe (46710), Pietro (46711), Antonio (46720) — which is a group, not four coincidences.

28 September — the register page is read, and the Austrian is a widow-shaped household of five.

L011_187.pdf is a scan with no text layer; the embedded JPEG was extracted and magnified. Matrícula dos Immigrantes entrados no Alojamento Provincial de Immigração em São Paulo, livro 011, page 187:

• 100927 — Franceschi, Francesca — «Mãi» — 37 — «c» — «Austria» — «Agricultor» — «Santos» — Maio 22
• 100928 — Giuseppe, filho, 14 · 100929 — Fortunato, 11 · 100930 — Maria, 9 · 100931 — Luigi, 6

Five people on consecutive numbers, and no husband on the page. A married woman of thirty-seven arriving with four children and entered as the mother of the party is the shape a widow makes, or a wife travelling ahead of a husband — the register does not say which.

All five are on the roster with estimated years only, because an age written by a clerk is not a birth date. The third child is Fortunato, not Vittorio — the first reading of the page at full-page scale said Vittorio, and magnifying the block corrected it before it was written down.

Still nothing connects them to this family. What the page adds is that the one Austrian entry is not a lone traveller but a household, which is a thing that can be searched for at the other end.

28 September 2026 — the newspaper row closes, and this item had it filed twice, two different ways.

⚠ It was not «never opened». The sources entry for dnc-nsk recorded a completed search of this portal, with findings — while this item carried the same portal as pending. The archive searched it, wrote it up under Sources, and left the worklist row open. A re-run then read the row, believed it, and reported a first opening.

dnc.nsk.hr answers on http:// and refuses https://. That is the whole reason this row sat «pending» — the archive had recorded the host and not the scheme, and two attempts on https came back as «denied or failed», which reads exactly like a dead site. A row that says «not opened» should say which URL was tried.

Opened, it gives fifty hits — 46 «Defranceschi», 4 «Defranceski», 1845 to 1920 — and three seams: the «Aquilone» seen from the Austro-Hungarian side with five new voyages including Cardiff in August 1876; a Defranceschi standing as the Italian candidate in the 1879 Istrian election and beating the Croatian dr. Klodić in the rural communes; and Antun Defranceski, an accounting councillor in the Zagreb administration from 1863. Full reading on /searched/.

Still open: three. Rijeka HR-DARI-382 kut. 1, Zadar's «Prijamni ured» and the Bakar parish registers. The «Stemmi della nobiltà di Buccari» and the newspaper portal are now closed, and São Paulo stays pending on purpose behind its bot protection.

All twenty «Primorac» sightings were read the same day, and they did roughly double the voyage table — from 22 rows to 42, and the West Indies run to Guadeloupe is now printed rather than inferred.

Handed over on the Senj line page, named and dated and not written up: her parents Ljubomir 1892–1966 and Milka 1886–1971, and six siblings — Leo 1920–2008, Karmela «Carmen» 1922–1987, Oto 1923–1945, Ljubomir 1929–2007, Mario 1933–1984, Anton.

What belongs here is the join, which is Hedviga. The rest is [the Blažević Archive's](https://daviddef.github.io/TheBlazevic/) ground, and a second family reconstructed on a page about a different surname is how an estate ends up with two versions of one person.

Walked: forty-nine nodes and 160 named people inside them, counted pen by pen and published as a table. Brazil 22, the United States 19, Italy 18, Croatia 16, Austria 14, Venezuela 10 — and Brazil being the largest is exactly what this archive had already written from the other direction.

None of the 160 is in the roster. Reading each of them properly — name by name, against the records — is the next thing, and it is a body of work rather than a row.

The row said «160 named individuals». It was a count of the folders. The pens hold 1,448 nodes, of which 1,054 carry a date: twenty-three country folders, each with town folders beneath it — Rijeka, Gračišće, Gologorica, Pula, Bakar, Sovinjak, Labin, Poreč, Kaštelir — and the people under those. They go back to a Nicolo Franceschi of 1318 and a Zadar line from 1436.

And they explain the 668. 486 of the «668 people only in the tree» are inside these pens — David has already tested and parked them, which is the opposite of untouched work. Correction 92.

What is left is 84, and it is the good list — Ližnjan, Crikvenica, Gračišće, Gologorica, Brooklyn, and two beside Pfullingen. It was announced as 113 for an hour, on a join that compared whole sets of given names and so reported the top of David's own male line as a stranger. He was in the roster all along. Correction 92.

So the job is no longer «read 160 strangers». It is import the 84, starting with the ones in parishes this archive has already read, and leave the pens as the judgement David already made.

Done, 20 September 2026 — the eighty-four are imported. 82 new roster rows; two of the eighty-four (Antonio De Franceschi, Maria Defranceschi) were already on the roster under the same name and were skipped rather than duplicated. The roster goes 1,225 → 1,307.

Twenty-five of them are living, and they are here by name and by town and by nothing else — no birth year, no death year, no photograph, exactly as the rule has stood since 15 September. The living gate passes on all of them.

Where they land: Ližnjan 7, Gračišće 6, Crikvenica 5, Brooklyn 5, Gologorica 4, Allentown 2, and singles at Trieste, Venice, Varaždin, Gallignana, Mercedes, Ellis Island, Urach, Metzingen, Horsham, Johannesburg and Castelgomberto. Thirty-nine have no place at all, which is the next job on them.

Every row says what it is: src: myheritage, and a note reading «Tree-derived: no register has been read for this person here». They are inf — the tree's own word, and nothing else yet. That is the honest state, and it is better than being invisible.

Done, and the answer is reassuring rather than startling: all seventy-nine are this family. Five collections.

3.1.1.1 (21 rows) is the same seven people under every spelling a clerk could manage — Lujza, Luiza, Lusza, Luise; Petar, Peter; Zdrafko, Zdravko, Ztravko. 2.1.3.1 (13) adds the town to each: Crikvenica, Fužine, Neuhausen, Urach, Metzingen, Reutlingen. 2.2.2.1, the War Time Card File — registration cards and employees' record books — holds Felix, Ivan, Peter and Zdravko. 2.3.3.1 has seven more including an Anton born 20 January 1890. 3.1.3.2 is the IRO period and, under its own heading, «Emigrations in 1949».

Three things the archive did not have. The CM/1 personal file, 3.2.1.1 ref 32110000 058.158, whose own title says «and of further persons» — now an errand. Ruža spelled HRELJEC, born 13 February 1922 at Fužine / Belnica, which is a date this archive did not have under any spelling. And a card with Ivan's name on it — David's grandfather, nineteen years old, in a German employment file.

The row said this was «one of the cheapest bodies of unread material the archive has». It was right, and it took an hour.

Done. Walked volume by volume from the fonds page on 17 September: Pola 1 · Triest AB 4 · Triest HB 3 · Görz 1, from 1848 to 1918, and every one of them opens without an account. They are a ninth provider in the builder and listed under the map with an address each — not on the markers, because the panel builds FamilySearch addresses and these are not FamilySearch, which is the same reason Antenati opens from its own page.

Two things had to be fixed to receive them. The record-type classifier knew five languages and not German, so Taufbuch, Taufen and Tote all fell into «Other» — it now reads them, and Gorizia's one volume is filed as a birth register rather than as something unclassifiable. And the «why is there no link» sentence had two branches and needed a third: a Matricula volume is not an unharvested address, it is an address somewhere else.

Nine households were added on 16 September so that the person pages could draw a family at all — Anton Rudolf and Ursula at Crikvenica, Zdravko and Ruža, Antony and Marion, Petrica and Michael, Ivan and Hedviga at Senj, Josip Petar and Božica, Petar Vid and Luise, and the two living generations at Johannesburg and Brisbane.

Every one of them is marked src: tree and says so on its own row: built from David's family tree and from the Arolsen file, the ship's manifest and the National Archives of Australia. Not one parish register has been read for any of them.

The books that would confirm them are already named and already ordered on the errands — Crikvenica births 1909–1926 for five of the children in the first household, and Senj 1926 and 1948 for the fifth. When those come back, these households stop being tree-shaped and start being read.

/map/ lifts its whole instrument into one fixed layer and asks for native fullscreen on top of it, and carries its entire view in the address bar — layers, basemap, find box, which filters are off, and where the map is looking. Both are written in AtlasTimelineTint.astro, which is this archive's own component and not the kit.

The per-provider maps under /map/<source>/ have neither, and neither does any other archive on the estate. A shareable map view is worth more to six archives than to one. If it earns its keep here it belongs in the kit — that is a conversation with whoever owns the shared UI, not a change to make quietly.

Eight rows on this list are stalled on Antenati returning 403 Forbidden, and they do not have to be. The tree cites Italia, Udine, Stato Civile (Archivio di Stato), 1806-1815 twenty-six times — Mione, Paluzza, Illegio, Naunina — and 22 of those 26 people are already in this archive's roster, read off Antenati's images of the same registers. Two sources worked apart, agreeing.

Which means FamilySearch has indexed the volumes Antenati will not serve. Collection 1939238 is already a provider here («fs-it»). Searching its index for the surname, comune by comune, reaches the entries without reaching the images — and the four Udine-cited people who are not in the roster say there is more in it than we have taken. Where the numbers come from.

Done, 20 September 2026 — and it opened the second door. The persona index takes q.anyPlace (f.recordPlace returns nothing; f.collectionId does not bind), so it was searched comune by comune and then swept across four hundred Udine records.

Nine records of the name in Carnia, eight of them at Paluzza and its hamlet Naunina, none of them in this archive. All nine are now on the roster with their ids.

The prize is Zuanne. «Daniele De Franceschi, 1782, Paluzza, son of Zuanne De Franceschi and Maria Delli Zolti» — and Daniele fathers Maria Anna at Paluzza in 1810 on Maria Silverio. Four generations at the place this archive has been calling its second door into the Val di Gorto and never opened.

What this does for the eight stalled rows: it proves the route. The images are still Antenati's and still 403 — but the entries are readable without them, and the parents come with the entry. Any Antenati-blocked item whose question is «who are the parents» can now be asked here first.

Not taken: the Francescone / Francesconi of Udine city, which is a different surname. Left open: Illegio, and whether «Soravito De Franceschi» of Ovaro is inside this study's scope.

Done, 17 September 2026 — and it was eight, not seven. The pen holds five households and eight baptisms in Santa Fe province between 1899 and 1918, with the surname in second position in three of them, which is why no search had ever returned them. Outside the pen it also produced Joso Defranceski of Ližnjan, dead at Buenos Aires in 1970 — the first Istrian of the name this project can place in Argentina — and one testable record against Rotilio. Where the name went now has Argentina.

Done the same day. 409 waypoints called, 409 arks, not one error.

Every volume in both collections is clickable now — 38 in Pola and Trieste across eleven Istrian and Triestine parishes, and 371 in Prekmurje and Međimurje across forty-seven towns filed under their Hungarian names. Čakovec as Csáktornya, Buje as Buie, Lendava as Alsólendva.

The trick that made it cheap: a FamilySearch waypoint id is «PREFIX:ancestor,ancestor,…» and the API answers on the prefix alone, so the whole job rode on a four-kilobyte list instead of a thirty-four kilobyte one. Paced at 2.2 seconds a call, which is the spacing that has never once drawn a throttle.

Four waypoints returned an ark another had already returned — not a fault. Muraszombat's district register covered several villages, was filmed once, and hangs under each of them.

No place on the map is now unlinked for want of an errand. The ones that still have no link have a reason: eight are Pazin catalogue entries with no images anywhere, nine are Antenati's and open from their own page.

David took a one-month full subscription on 16 September 2026. Every previous pass ran against a clock: the 9 September trial and the one that lapsed on 14 September are both on the search register, sixteen rows between them, and both end with a note about reading transcripts before the trial died.

What is already in: all 38 record sets the surname touches, paged out under three spellings; 314 movement records (221 published); 153 newspaper articles; the American naturalisation files and census images; the Victorian probate references.

What the clock stopped. Transcripts at scale rather than one at a time — an index row gives a name and a year, a transcript gives parents, ship, street and occupation. The images behind them, which is where a signature or a birthplace actually is. The Irish, Scottish and English parish sets never opened. And the UK Electoral Registers, which hold living people and can be counted here but never published with a place or a year.

How to drive it is written down: it is a client-rendered SPA, the results URL wants «_page» and not «page», and its own pagination anchors must be clicked or the harvest loses its state. The session is in the in-app browser, on findmypast.com.au this time rather than the .co.uk domain.

21 September — the three national sets are open.

Read on the search register: Ireland is a different surname entirely (11 rows, all de France, Defrance, de Frances — the Huguenot name), Scotland is one probate row and not ours, and England and Wales on the exact spelling gives five, of which one is Emerico's known 1876 Cardiff marriage, three are living people named here without detail, and one is new: Laura Defranceschi, born about 1848, died Q4 1931 at Croydon aged 83, GRO 2A/520 — the only person of this surname in the whole England and Wales death index.

What remains on this item is the UK Electoral Registers, which hold living people and can be counted but never published with a place or a year.

Done the same day. The map has a fourth switch: the records, 156 of them on 31 places, in amber.

It is a different kind of claim from the other three and it is drawn that way. The shelf, our people and the ground all say where this family is. A census line or a landing says somebody of the name was at this place on one day — so the panel says «named in a record here», never «who the archive puts here», and a place with both says both. 21 records name only a state or a country and get no dot at all, which is the same rule that keeps «Americas» off the map.

The work was mostly in the gazetteer, which had been throwing away the region a place string names: San Lorenzo, Puerto Rico had landed in California. What it cost.

Two of the fourteen were never disagreements. An Antonius buried at Pula as an infant in 1893 had been matched to the Antonius of Gologorica who died in 1893 aged seventy-nine, because they share a year. Twenty of this archive's burial rows read b = d, a child born and buried in the same year. The anchor test refuses that pairing now and the correction is logged.

The twelve that survive are three different problems.

Seven are age arithmetic — death agreed, birth out by one: Josephus, Cattarina, Jacobus, Valentinus, Davide, and Albino and Maria out by two and three. A record that gives an age rather than a date produces a birth year that depends on whether the birthday had passed. Not worth a morning each; worth a sentence on each page.

Two are too big for arithmetic: Mario De Franceski 1919 against 1924, and Girolamo of Omiš 1632 against 1622.

Three dispute the death year, and that is the serious kind: Clementina Maria of Gologorica, 1840 here against 1870 — thirty years; Josephus Lambertus, same parish, twenty-one; Lina of Torrington, two. The Gologorica registers settle two of the three in an afternoon, and this archive has already read them once.

Done, 20 September 2026 — each disagreement now says so on the person's own page. Fourteen rows annotated, one per disagreement, matched on name and source rather than on name alone.

That distinction cost a first attempt. Matching on name put the note on 56 rows, because this roster is record-level and Maria Defranceschi alone is twenty-four rows — the annotation would have accused twenty-three innocent namesakes of a disagreement that belongs to one of them. Stripped and redone.

The three kinds are written differently, because they are not the same problem. The seven off-by-one-to-three births say plainly that an age produces this and that the archive keeps its own figure. The two too-big-for-arithmetic say one side has the wrong man. The three disputed death years say this needs the burial register, not a judgement. And the mirrored Petrus Defranceschi pair says what it is: one man on the roster twice, 1768 and 1765, same death year 1829, at Gologorica and at Rovinj, standing until somebody resolves them.

Maria Defranceschi of Gologorica is merged. Two rows, born 1786 and 1789, dying the same year at the same parish under two MyHeritage identifiers. Both identifiers are kept on the surviving row; the birth year is not settled and the Gologorica baptisms will decide it. Correction 94.

Petrus Defranceschi is not merged, and that is the finding. His two rows disagree on the parish: a Petrus of Rovinj resting on a burial register with an ark, and a Petrus of Gologorica resting on the 2018 family file. Three years and a hundred kilometres apart is either two men or one badly recorded, and a burial entry outranks a tree entry. Both rows now carry a note pointing at the other. The Gologorica register decides it, not a merge.

The list is made and the door is half open. Of the 405 arks David's own citations name, 328 appear nowhere in this repository: 234 indexed records and 94 film images.

The session is signed out where it matters. Tested on 16 September: the waypoint API still answers 200 — which is why the Međimurje and Pola harvest ran this morning — but every route to a record returns 401 Unauthorized: the persona endpoint, the ark URL itself, and the search service in its place. A waypoint is public; a record is not.

This archive does not sign in on David's behalf, so the row is his: sign in to FamilySearch in the in-app browser, and a machine can read all 234 indexed records in one pass at the spacing that has never drawn a throttle.

The 94 image arks are a different job again — an image is a page, and a page is read by eye. Those are the ones worth doing first whatever happens: each is a register page David has already looked at.

22 September 2026 — David signed in, and the door is open. Tested the same minute: https://www.familysearch.org/ark:/61903/1:1:X9DP-9D17 now renders the whole record — name, sex, birth date, event type, event date, event place, both parents and the citation. «Antonio De Franceschi · Birth Aug 1704 · Baptism Nov 1704 · Croazia, Austria · Father Mattio De Franceschi · Mother Zuanna.»

One caveat that changes the method. The JSON routes are still shut — /platform/records/personas/<ark> and the ark with ?_format=json both return 401, because they want an OAuth bearer token and a cookie is not one. The rendered page is the way in, one navigation at a time, which is slower but entirely automatic. 234 indexed records is a long afternoon, not a blocked row.

This row is now next rather than blocked, and it is the largest piece of work on this list that needs nobody.

22 September, an hour later — the number was wrong, and it is ten.

Rebuilding the list from the two files in this repository that hold David's own citations — myheritage-surname-2026-09.tsv and to-the-tree.ged, 351 arks between them — and diffing against every ark the site already carries leaves ten, not 328. The week's sweeps absorbed the rest.

Two traps in the counting, both of which inflate it. The repository stores some arks bare, without the 1:1: prefix, in an arks array — a regex that looks for the prefix reports them missing when they are already here. And data/fs-catalogue.json holds 5,451 arks which are first images of catalogue volumes, not citations at all; counting those turns ten into thousands.

What the ten gave. Three are indexed records: Vittoria Defranceschi, born 29 April 1894 at Telve, daughter of Antonio and Rosa Mattivi; Estefania Defranceschi, married at Dallas on 24 July 2004 to Lance C Thomas; and one that is restricted to certain accounts and cannot be opened. A fourth, Marina De Franceschi 1710, turned out to be here already under a reading — and the index adds her mother, Vittoria, whom the page did not name. Seven are film images and are a read rather than a fetch.

The original figure was honestly arrived at and is honestly withdrawn. A count of what is missing has to be made against every way the thing is stored, not one of them.

22 September — done, and it was the seven images that paid.

Three households in three towns, none of the four films referenced anywhere else in this archive: a tailor's house at Kaštelir with three of the name in it and a grandfather named Giulio; a widow of eighty-five at Bakar, born about 1734, with an unnamed husband behind her; and a carpenter's family at Parenzo whose son had the Marchese de Polesini for a godfather. The read.

The row is closed. What remains of the arks is one record the owner has restricted and the right-hand page of one image, and both are named rather than glossed.

The file is written: data/to-the-tree.ged, 494 people. It is not an edit to David's tree and it is not going to be one — a GEDCOM is a file he can read before importing and undo after.

Why 494 and not 780. The 780 counts roster rows the tree has under no name-and-date match. This file is stricter: it leaves out anybody whose given names already appear in the tree at all, even with different dates, because pushing those in makes duplicates rather than fixing anything. 691 were skipped on that test.

Where they came from: 168 out of the 2018 family file, 70 from households read off the page, 54 from name indexes, 46 from headstones, 42 from volumes read end to end, 41 from indexed records, 25 from burial registers, 14 from the Gologorica books. Each carries a NOTE saying which, 223 carry a FamilySearch ark, and every one links back to its page here.

No living person is in it. Twenty were dropped for being flagged living and one more for being born after 1930 with no death recorded — a tree entry is a date by construction, so the rule here is stricter than the name-and-town rule the rest of the site keeps.

22 September 2026 — it cannot be regenerated, and the reason is exactly right.

export-to-tree.py needs the parsed live tree to know who to leave out — its whole discipline is «leave out anybody whose given names already appear in the tree at all», and it cannot apply that against a tree it cannot see. Run today it stops with «Pass the parsed gedcom.json so the tree side can be matched».

The tree export is not on disk and was never meant to be. read-gedcom.py says so in its own header: «The GEDCOM itself is NOT committed. It carries dates of birth for living people.» The copy David exported on 16 September has gone from the downloads folder, and that is the correct state for a file full of living people's birth dates — it is not a bug to fix.

So the file offered back to the tree is frozen at 17 September, 494 people, while the roster has grown to 1,419 rows — and the growth is exactly the kind this file exists to hand over: a tailor's household at Kaštelir, a widow at Bakar born about 1734, a carpenter's family at Parenzo, four households inside the Svetvinčenat years no index covers.

The remedy, in order, and it starts with David: export the tree again → python3 scripts/read-gedcom.py <file.ged> --out <dir> → python3 scripts/export-to-tree.py <dir>/gedcom.json. Ten minutes of machine time behind one export. Moved from «next» to «blocked», because it was listed as work this archive could do alone and it is not.

It existed all along. Eight reports came off this account on 15 September and only five were opened; Ivan's was one of the three still unread in the inbox. Found and read on 16 September.

94 matches in thirteen clusters, and the first holds fifty people — inside it, David's entire cluster of twelve plus Račić from his cluster of four. From the father those are one cluster, not three.

Every cluster David has is now decided by arithmetic. Clusters 2, 5 and 6 have ten, three and two members in his father's report and none in his mother's; clusters 1, 3, 4, 7 and 8 are the other way round. No cluster has members in both. Cluster 9 is in neither and is still called unplaced.

The table. This morning the Croatian three were an inference from a pool of surnames. They are proved.

All ten opened on 16 September. They do not name a common couple, and the reason is that eight of the ten are shallow — three of them show three, seven and twelve people. The two deep ones, at 3,597 people and 1,521, do not touch each other on a single named person.

What they share is a surname pool, and one name carries further than the rest: Vukelić stands in three independent trees. Prpić stands in two. Behind them Blažević, Pavelić, Krmpotić, Tomljanović, Biondić, Lopac, Katalinić, Rončević, Šojat, Butković, Šolić, Delale — every one a Senj-hinterland Bunjevac name.

So the next name to chase is Vukelić, not Blažević. Blažević is the one this family carries and therefore the one the eye goes to; Vukelić is the one that recurs across the trees with nothing else in common. Written up.

The sweep of 16 September read ten transcripts out of 424 publishable records, and every one of the ten paid: a Cosulich bride at Cardiff, a GRO volume and page, three Italian sets closed as Neapolitan.

An index row gives a name and a year. A transcript gives parents, a parish, a ship, a street — and for the Italian sets, a note in Italian naming the frazione and the grandfather. The transcript pages are fully client-rendered with no server-side payload, so there is no shortcut: they come one navigation at a time, about five seconds each. Start with the 277 United States passenger records, whose transcripts name the town of last residence — which is the question this archive actually has: which of these six hundred are ours?

21 September — counted, and the order of work is inverted.

All sixteen pages harvested: 307 records, and they are 270 passenger and crew lists, 13 naturalisation petitions, 11 Italians to America, 6 Canadian border crossings, 2 Illinois naturalisation index, 2 passport applications, 2 British & Irish Roots, 1 UK outbound. The count is on the search register.

The naturalisation petitions carry «Birth place as transcribed» in the transcript itself. No image, no order, one navigation — and that field is the question this item exists to answer. Thirteen records answer it outright; two hundred and seventy need a plate each. So the passenger lists are not the place to start.

All thirteen were read, and all three households were already here — Joseph at «Gallignana, Austria» (the first of the Eleven to go), Peter at «Basel, Switzerland», and Bruno, Gerda and Richard at Brooklyn under the alias «de Franceschi or Seedorf». That is a real result: the naturalisation route is exhausted and it added nobody.

What is left is the 270 passenger records, one plate at a time, and they should be attacked by arrival year rather than alphabetically — the Istrian and Kvarner departures cluster in 1900–1914 and again in 1949–1956.

Two traps, paid for. A full page load in this SPA wipes window, so a harvest must be dumped out of the browser before navigating anywhere — 307 rows were collected and then lost to a single transcript click. And the Apollo cache (window.__APOLLO_CLIENT__.cache.extract()) holds index fields only: Record: keys carry what the results table shows and FulfillableItem: carries nothing but a price, so there is no bulk shortcut to transcripts. The subscription does cover them — remainingLegacyCredits is null and isPurchased is true — so nothing is being spent.

Struck, and the reason is that it was already done by a better source.

The transcripts do carry Residence with no image needed — that part of the plan was right. But the Ellis Island sweep already read 1,400 arrivals and 429 recorded residences, and found exactly two Istrian in the whole database. Findmypast's 270 are a subset of the same arrivals.

And there is no query that shortens the job: keywordsplace matches ports, not residence — Domenico's transcript says «Residence Ponte Alpi» and searching «ponte alpi» returns zero, so «Fiume: 0» and «Istria: 0» are worthless as negatives.

What survives is narrower and worth doing: the 1949–1956 arrivals, which the Ellis Island database does not reach, and where the Istrian esodo sits. That is a new item, not this one.

One hit in the whole of them. Of the 44 record sets the surname touches, the only British parish register holding anybody of the name is Wales Roman Catholic Marriages, with the single 1876 entry that started all of this. No Irish, Scottish or English parish register has one.

For a Catholic family of the Adriatic that is what one would expect. It is written down so that nobody opens them again. The register.

Nobody. The sweep paged out every result for four spellings — 644 records, 44 record sets — and neither collection is among them.

That is a finding. The South Wales Echo called them «an old Cardiff family» in 1896 and a De Franceschi married a Cosulich there in 1876. By 1921 there is nobody of the name left in England or Wales. The register.

Done, and it ends the question. Two of the San Lorenzo women were read in full from the 1920 census: Cármen Colón de Franceschi, 23, and Aurora Cruz de Franceschi, 25. Both born in Puerto Rico to parents born in Puerto Rico, both entered as esposa, both recorded by the enumerator as Black.

An Afro-Puerto Rican family two generations on the island by 1920, married into a surname that reached the Caribbean with the nineteenth-century Corsican settlement. Not immigrants, and not ours. Written up.

The search is right and the door is shut. Since 16 September this family has a Cosulich in it — Emerico married Marietta Cosulich at Cardiff on 26 July 1876 — and the Cosulich were shipowners of Lussinpiccolo and Trieste. A Cosulich ship at Cardiff in the 1870s, printed beside a De Franceschi master, is the join this archive is missing, and the shipping columns are where it would be printed.

Both routes are behind a bot check. The British Newspaper Archive answers a security challenge before it will show a result; Findmypast's own newspaper search carries a reCAPTCHA field on the form. This archive does not work round bot checks, so the search waits for a person.

What to search, when you have five minutes: Cosulich on its own, then Cosulich with Cardiff, then Cosulich with Aquilone. The 153 articles already read under this family's own name are on the search register; the earliest sixty of them are shipping intelligence, which is exactly the column a Cosulich would appear in.

28 September 2026 — a third route, open and free, and it half-answers the question.

Both routes on this item are behind bot checks. A third is not. The [Portal Stare hrvatske novine](http://dnc.nsk.hr/Newspapers/Search.aspx) carries the Austro-Hungarian shipping columns of «Primorac», answers on http://, and has no challenge of any kind. This archive did not work round anything; it used a different source.

What it gives, and what it does not.

• Cosulich → 81 hits — but they cluster 1900–1925: «Naša sloga» 1901, «Jutarnji list» 1912 and 1914, «Sušački novi list» 1925, «Hrvat» 1904. Not the 1870s shipping intelligence this item wants.
• Cosulich Aquilone → 0. No page on the portal carries both names.
• Aquilone → 32, of which 24 are now read — twenty naming Defranceschi as master, four more from 1874–75 where the column prints no master at all.

⚠ So the join is not made here. The question — a Cosulich ship at Cardiff printed beside a De Franceschi master — is not answered by the Croatian press, and the British half still needs a person. But the item can stop saying the search is impossible: one of its three routes is open, and it has been run.

Eight «Aquilone» hits remain unread on page 2 and beyond.

Done. Antonio De Franceski, born 1875 in Italy, arrived 1902, in the 19th Ward of Memphis by 1910 with Bertha born in Germany and nine children born 1904–1917 — Pietro, Aurora, Alba, Albert, Flora, Estelle, Serena, Juliet, Romeo and a second Anthony. Three censuses, twenty-one records, and the 1910 sheet spells them «de Franceski».

Born in Italy in 1875 and leaving in 1902 is the Carnian window. Istria in 1875 was Austria and an American enumerator would have written Austria. It is not proof and the census says no more than «Italy» — his naturalisation petition would name the town. The reading.

Done, and the residue is published rather than hidden. GeoNames files a village too small for its own settlement record under its church or its railway stop, and this builder was reading only populated places. Reading churches, monasteries, hospitals, stations and localities — and nothing else — put four more on the ground: Mađarevo is a railway stop, Mandalina a halt, Ludina a station, Komarevo a locality.

A stream or a hill of the same name is not the village and is deliberately left out. The Glogovnica in the dumps is a river; taking it would have been a wrong dot rather than a missing one.

Forty-one places carrying 130 volumes still cannot be drawn, and the map page now names them instead of letting them be invisible. Three are not settlements at all — Bolnica Dubrovnik, Nahodište Dubrovnik and Nahodište Split are a hospital and two foundling houses, and should in the end hang on their towns. The rest need a Croatian source and a person, not another pass of the same script.

Done. scripts/place-audit.py applies the two tests the builders already had — a place catalogued only by a provider that works one region cannot be outside it, and a place whose own name carries a country cannot be in another one.

It found six. Senja, Sv. Juraj — the Senj deanery's coastal parish, eight kilometres from the town Hedviga Blažević was born in — was outside Belgrade. Sv. Jakov-Šiljevica, the old name of Jadranovo four kilometres from Crikvenica, was in Šumadija, and so was Klasnić of the Banovina. Krug was in Lower Austria. Grad and Tišina, both Prekmurje parishes, were in Dalmatia.

Three faults underneath them: the region box covered only the two Udine providers and now covers all seven; «Sv.» was stripped and never expanded, so GeoNames' Sveti Jakov Šiljevica and the catalogue's Sv. Jakov-Šiljevica never met; and this builder had never read data/gazetteer.json, where a hand-checked coordinate lives with its reason beside it — so a hand fix could not reach the research map at all.

The audit reports zero. It changes nothing itself and should be run whenever a place is added. Correction 91.

Done. The sweep page now carries every publishable record: 177 standing on a place and 247 whose place is a country or nothing, in two folded lists. A reader can look for their own grandmother instead of being told how many there are.

The 220 withheld are still withheld — electoral registers, Social Security index, obituary notices, divorce indexes.

Two names in the unplaced list are worth somebody's eye. Bosiljka and Josip Defranceski, born 1918 and 1915, crossing in 1954 — Croatian given names under this family's own fused spelling, and a post-war date. And a Marta Defranceschi, 1922, which is the given name the Pićan errand has been trying to settle for a fortnight.

Done, and it found none. scripts/living-audit.py reads the built HTML rather than the sources, because the built HTML is the only thing a reader ever sees: 37 people the archive holds as living, against 1,280 pages, looking for a four-digit year standing in the same run of text as a name.

Getting to none took three narrowings, and each is a lesson. The menu lists every page title on every page and one of those titles is a person's name, so the audit reads only what follows the first heading. No tag may stand between the name and the year — without that a households row hands the next person's birth year to the one before. And no sentence may end between them, because «David Defranceschi. The 1584 plaque…» is prose about a plaque.

What is left is the search index, where the page title «1948» sits beside a name in a lowercased string. On the method page: a rule enforced at the door should be checked at the window.

The row promised thirteen birthplaces. The transcripts do not carry one.

Four were read on 16 September — Giovanni of Chicago (naturalised 1937), Giovanni of Connecticut (passport 1922), Louis (passport 1919) and Tobias de Franceski (passport 1922). Every one of them gives «Birth place: Italy» and nothing finer. Tobias is the exception and only because he was born in the District of Columbia.

The town is on the image, not in the index. Findmypast holds the images and links them from each transcript, and the transcripts carry the NARA reference that would fetch the same page from Washington — M1490 roll 1807, roll 1008, roll 1842, M1285 roll 41.

So the work is a different job from the one written down: eighteen images to read by eye, not eighteen transcripts to harvest. Two are already an errand because a petition is worth ordering whether or not the image reads.

Corrected the same day it was written, which is the right speed for a wrong plan.

21 September 2026 — the images are open, and two towns are out.

The viewer was never the obstacle. A Findmypast transcript id for these sets is a FamilySearch film and frame — US_PASSPORT_APPLICATIONS/**007572652**/**00658**/… — so the plate can be fetched directly and read on a canvas, with no viewer, no filmstrip and no paging. The route is written down.

Louis De Franceschi was born at Sequals, Province of Udine, on 3 November 1873; father Pietro, dead by 1919; Havre to Washington in February 1907; naturalised 4 October 1915; cement decorator, which is the Sequals trade. Giovanni DeFranceschi was born at Castronovo on 10 December 1876; father Francesco, dead by 1922; Havre to Torrington, Connecticut in April 1903; naturalised at Winsted in 1911.

And the index was never wrong. The form asks the father's birthplace in one blank labelled «(Country.)», so «Italy» is the whole of that field. The town lives on the applicant's own line, and only on the naturalized-citizen form — which is why Tobias's native-citizen application of 1922 has no town to give, and his father's has Sequals.

Still to read: the remaining images, and the Chicago naturalisation of 1937, which is a different series with a different image path.

27 September — the naturalisation series read, and «they all say Italy» was true of four records, not of the set.

The 16 September reading took four passport applications and found «Birth place: Italy» in every one, correctly, because that form asks the father's birthplace in a blank labelled «(Country.)». The naturalisation petitions are a different form and they carry the town. Ten were read today off the transcript, without an image.

The towns that came out:

• Joseph Defranceschi — «Gallignana, Austria», born 12 March 1892. That is Gračišće, the parish this family comes out of. Declaration 22660 (Brooklyn, 1912) spells it «Gaglinana»; petition 13838, certificate 1104860 (1915) spells it «Gallignana» — the same clerk's office, three years apart, and the second is right.
• Gerda De Franceschi — «Hollen, Germany», born 15 January 1913. This archive held «1913» off a gravestone and no birthplace at all.
• Peter De Franceschi — «Basel, Switzerland», born 12 January 1902. A country this archive has never recorded for anyone of the name.
• Giovanni of Chicago, 1937 — «Italy», and nothing finer. Checked, and the earlier reading stands: NARA M1285 roll 41 is a Soundex index card, not a petition, so there is no town on it to find.

And one document gives a whole household. Declaration of intention 284221, certificate 2-665402, Brooklyn 1939, names Bruno (b. 12 Aug 1907), his wife Gerda (b. 15 Jan 1913) and their son Richard (b. 2 Dec 1932) on one paper. That is a source naming a couple together, so it is the first spouse link written from a record rather than from the direct-line chart. Gerda's oath of allegiance follows in 1944, certificate 5742645.

Richard is added as a new person and deliberately not joined to the Richard Defranceschi already on the roster from David's tree, who carries no dates: nothing links them but a forename.

She appears once in the whole of Findmypast and it is the consular marriage itself: GRO MCON, volume 788, line 70. The index does not name the groom, and the finder offers Helen Whittall, Frederick P Wilkinson, Reginald E Turrell from the same page.

Unresolved, and left so. Only a consular certificate would name her father, which is the one thing that could join her to anybody here. Written up.

Done. Giovanni «John» De Franceschi, born 1895 in Italy: the 1930 Chicago census with Freedia and the children Thersa and Ateo; naturalised at Chicago in 1937 (NARA M1285, roll 41); buried 12 April 1951 at St Joseph's Cemetery, River Grove. Frieda died in 1988 and Attilio F Defranceschi, 1925–2010, is almost certainly Ateo.

Every record says «Italy» and stops. The petition behind the Soundex index names the comune and is now an errand.

Opened on 16 September. 96 matches in six clusters.

It shares 31 matches with David's mother and 6 with David — five to one, which is what a relative on her side looks like, because a child inherits about half of what a parent has.

It splits her fourteen clusters nine to five. Nine contain somebody the third kit also matches; five contain nobody. 81 of her 103 matches are in a shared cluster, 22 are not. The nine are her mother's family; the five are her father's — and the father's side belongs to another session, so this archive stops there.

The by-product is what matters here. Against David's own nine clusters the third kit touches 1, 4 and 8 — and not 2, 5 or 6, which are the Croatian three. A maternal relative, tested independently, lands in three of his clusters and in none of the Croatian ones. That is arithmetic on a third person's DNA rather than an inference from a surname.

Whose kit it is is not published, for the same reason no match's name is.

Three documents this archive cannot read from a desk. Two were found on 16 September and answer the same question — where in Italy — and the third, found on 17 September, answers a different one: who was in the house.

The Cardiff marriage certificate — GRO, Cardiff district, Q3 1876, volume 11A, page 428. The parish register gave the two fathers and nothing else; the certificate gives both ages, both residences, both occupations and the witnesses. If Emerico's occupation is master mariner, the at-sea page stops being a collection of sightings. £12.50, ordered online.

Giovanni's Chicago naturalisation petition — NARA M1285, roll 41. Every American record for that household says «Italy» and stops: the 1930 census, the burial at River Grove in 1951, the Soundex index itself. The petition names the comune. Antonio of Memphis, born 1875 and arrived 1902, has one too and it is worth asking for in the same letter.

Both are on the errands page in full.

The Arolsen CM/1 file — 3.2.1.1, ref 32110000 058.158, filed under Petar Defranceski, born 15 June 1913 at Crikvenica, and titled in Arolsen's own words «and of further persons». A CM/1 file says where a displaced person was during the war, who was with them, and where they asked to go. Arolsen publishes online without an account and answers enquiries free. Where it came from.

David's tree gives Nelida Albina de Franceschi, born 9 September 1931 at Mercedes, as Rotilio's daughter, and cites Argentina, Buenos Aires, registros parroquiales, 1635–1981, ark 6ZP4-29WC. The citation reads «Entry for Nélida Albina and Federico Nebbia, 30 de noviembre de 1931» — and in every other citation from these Argentine collections the second name is the father. On the face of it the record is a Nebbia baptism.

It is the only Argentine record ever attached to Rotilio's story, so it decides whether Argentina stays dotted. Signed out, FamilySearch answers 401 on the ark — tested 17 September. It needs one click in David's own session.

22 September 2026 — signed in, and the answer is that the account cannot see it. The ark no longer returns 401; it renders, and what it renders is:

«We're unable to show this record. This record can only be displayed on certain accounts. FamilySearch must honor the agreements we have with our partners, record owners, and internal policies.»

So this was never a sign-in problem. The Argentine parish collection is restricted by its owner, and David's own account is not one of the accounts that may see it. A second sign-in will not fix it and should not be attempted.

The question stands exactly where it was: the citation names Nélida Albina and Federico Nebbia, and in every other citation from these collections the second name is the father — so on its face the record is a Nebbia baptism, and the only Argentine record ever attached to Rotilio may not be his at all.

What would answer it now: the image rather than the index entry, or the parish register at Mercedes (Buenos Aires) direct. Keep Argentina dotted, not solid.

The register already read gives «Rutilius 3-11-1902, married 16-5-1928» and this archive took the wife's name from the 2012 reunion album — which turned out to be his sister's husband's surname. David's tree still carries an «Albine Chinkopan, born Italy, died Buenos Aires» against that same date.

The register itself names the bride. One entry, one date, one parish, and it settles a question two pages are currently hedging. Croatian registers of 1928 fall inside the hundred-year closure, so this may be an errand rather than a click.

21 September 2026 — ANSWERED, and the answer reverses this archive's own reading. FamilySearch full-text, query Rotilio Defranceschi, reached Trieste, Marriage Proclamation Records 1928, Parte II Serie A n. 132: «Defranceschi Rotilo … figlio di fu Antonio … e di Valhart Maria» marrying «Cincopan Albina … figlia di Giuseppe … e di Trompich Maria». He did marry an Albina Cincopan. The album was right; the correction published here was wrong. Three people added, a trade recorded, and Trieste added as a place the bride had lived. See /corrections/.

The fourteen children were read off the Liber Baptizatorum of St Vitus on 6 September and no ark and no image number was kept for any of them. Ten are also in the FamilySearch index and can be re-joined by their household arks; four — Ignatius Francus Mattheus 1745, Vitus Joannes Baptista 1751, Aloysius Rochus 1754, Rosalia Francisca 1755 — are in no index at all, so their pages rest on a reading nobody can check. This archive's own rule is that a read is worth what its citation is worth.

20 Sept 2026 — eleven of fifteen recovered, and the list turned out to be wrong in two places. Matching the index on the parents rather than the children (q.fatherSurname across Defranceschi / De Franceschi / Franceschi, keeping only rows with a Valentin- in them) returns eleven rows, six of them this household's: Matheus Franciscus 1739, Aloysius Joannes Petrus 1752, and the burials of Mattio 1742, Antonio 1753 and Aloysia 1757 — plus Jacobus Vincentius, baptised 21 July 1757, who was not in the list at all. His entry was read on the page before it was believed, and with him Valentino's sons come out at ten, which is the number the family has always given. Reading the Aloysia frame settled a second thing: the entry is a burial and gives her age as «d'anni 4», so she was born about 1753, not 1757 — which is just as well, because no mother bears a live child eight weeks after 20 July.

What is left. Four children are in no index — Ignatius Francus Mattheus 1745, Vitus Joannes Baptista 1751, Aloysius Rochus 1754, Rosalia Francisca 1755 — and four more come back on no spelling tried: Josepha Catharina 1737, Nicolaus Francus 1741, Nicolo 1742, Francisca Joanna Theresia 1742, Maria 1757. All of them need the St Vitus film read at their dates. The ark prefixes are the map for that: …-W… frames are the baptism book, …-QD… frames the burials.

Later the same day, the remaining gap closed to four. Filtering the surname search on a date range (1735–1760) instead of on the father's name returns twenty-two rows, and picks up five more of the household: Josepha Catharina 1737, Nicolaus Francus 1741, Francisca Joanna Theresia 1742, and the burials of Nicolo 1742 and Maria 1757. Eleven of the fifteen now cite an image. The four that never come back on any spelling or filter are Ignatius Francus Mattheus 1745, Vitus Joannes Baptista 1751, Aloysius Rochus 1754, Rosalia Francisca 1755 — exactly the four this archive named as unindexed on 6 September, which is as good a check on that reading as the index can give. Those four need the St Vitus film at their dates, and that is all this item has left in it.

Closed the same day — all four were read on the film. Any ark resolves to a film and frame number through /service/records/storage/dascloud/das/v2/3:1:<ARK>, so the eleven indexed children gave eleven dated frames to interpolate between. Ignatius Francus Mattheus is frame 126, Vitus Joannes Baptista 228, Aloysius Rochus 257, Rosalia Francisca 272 — and every one sat where the 6 September reading said it would. Three of them gave up a birth date that reading had missed: 28 July 1745, 16 August 1754, 3 October 1755. All fifteen children now cite an image, and this item has nothing left in it.

Every sweep this archive has run searched the SURNAME. Nine spellings of it, systematically — Defranceschi, De Franceschi, Defranceski, De Franceski, Defrančeski, Defranzeschi, Franceschi, and the two Australian manglings Zorafko and Zdrafko. That is why the Australian and Arolsen files were found.

The given name has never been varied once. Out of 312 rows in the research log, eight mention searching by a forename and all eight are a hunt for one named person, not a sweep.

That is a hole with a shape. Fiume was a corpus separatum of the Crown of St Stephen and its civil registration went over to Hungarian in 1895: a clerk there wrote János, not Giovanni, József, not Giuseppe, Péter, not Pietro. Hungaricana is free and this archive already uses it. The same for the Austrian sets and Johann, Josef, Franz, Andreas. Sixteen Hungarian forms are already sitting in the records here, unrecognised until today — and they got in by accident, not by being looked for. The seven columns are a search plan that has never been run.

Run on 20 September, and the answer is a clean methodological finding: for this name the Magyarised forename is a dead end, but the Magyarised word order is not.

Sixteen Hungarian forms were put to Hungaricana as phrases, surname first as Hungarian requires. Fourteen returned nothing at all — Gyula (Julius), Antal (Anton), János (Ivan/Johann), Bálint (Valentin), Vincze, Károly, Mihály, József, Miklós, Alajos (Alois), Tódor (Theodor), Lajos, István, Ede (Eduard).

Two returned, and the pair explains the other fourteen. «Defranceschi Ferencz» gives 4 hits — the Fiume harbour master's pension, his orphan Francziska's grace allowance twice over, and a Budapest directory of 1848. «Defranceschi Péter» gives 1. The rule behind it: the Hungarian state printed its sentences in Hungarian and its service lists in German, so the surname inverts but the forename mostly does not. Ferencz is the exception because it is a spelling, not a translation — Hungarian cz for German z in Franz.

The control matters more than the zeros. Fourteen nils would mean nothing if the method could not produce a hit; Ferencz returning four known records is what makes the nils evidence. The sweep was run because the search log showed 312 rows in which the surname had been varied nine ways and the forename never once.

And it was not empty. Péter turned up a fourteenth man for the imperial roll — «beim Husz. Reg. Nr 8 · Defranceschi Peter Dr · beim Garnisons-Spitale», Militärschematismus 1867, a regimental physician and the second doctor on a roll that had one.

Two things are now settled and the third is the whole job.

Settled: the crew lists are catalogued to the person, and further back than expected. «Ship: Belgravian; Official number: 98611. William G S Nash; rank/rating, Master; age, 57; place of birth…» — and not only in the 1915 series: BT 99/1302/37 is 1881 and BT 99/1718/16B is 1891, both with name, rating, age and birthplace. This archive's first explanation of the blank — that shipping is catalogued by vessel and year rather than by master — was simply wrong, and is retracted.

Settled: the name is not in it. Franceschi rank rating 0, Defranceschi seaman 0, against 827 for «place of birth, Italy» master. The index is dense, so the absence means something.

Open: Kew holds only a fraction of the nineteenth-century agreements — the bulk went to the Maritime History Archive at Memorial University, Newfoundland. Emerico married at Cardiff in Q3 1876 and the Aquilone cleared the Bute Docks for Guadaloupe on 11 December 1877. Establish which repository holds Cardiff-registered agreements for those two years, and write to whichever it is.

Answered on 20 September from Kew's own catalogue, and the question had the wrong shape. It is not «Kew or Newfoundland». It is three repositories and one number.

• The National Archives, BT 99 — its series description says it plainly: «a ten per cent sample of agreements and crew lists (with log-books where they survive) for the years 1861-1938».
• Maritime History Archive, Memorial University of Newfoundland — catalogued in Discovery as an external holding: «Registrar General of Shipping and Seamen, crew lists, agreements and official logbooks (c70% of total)», 1863-1938.
• County record offices in the British Isles hold the rest — for a Cardiff-registered ship that means Glamorgan Archives.

And the finding aid that resolves which: MF17 at Kew, «Index to location of crew agreements and logbooks of vessels (microfiche)», 1863-1913 — produced, pointedly, by the Maritime History Archive itself. Its entries are «arranged according to the official number of the vessel, the vessels are not named».

So the chain is fixed and its first link is missing. Ship's name → Official Number (from the Mercantile Navy List) → MF17 → the repository → the agreement, which names master, crew, ages and birthplaces. Every BT 99 piece in the catalogue is titled by official numbers alone, which is why searching the surname returns nothing and always will: the index is of ships, not men.

Which makes the Cardiff marriage certificate the hinge of the whole maritime thread. GRO Cardiff, Q3 1876, vol. 11A p. 428. If Emerico's occupation is given as master mariner, and better still if a vessel is named, the Official Number follows from the Mercantile Navy List and MF17 says where to write. Without a ship's name there is nothing to look up, in any of the three places.

«An old Cardiff family» is a phrase a newspaper used in 1896 and this archive has never tested against a census. Findmypast's 1921 Census and 1939 Register hold nobody of the name, which dates the disappearance but not the family. The censuses between Emerico's marriage in 1876 and 1911 have not been searched at all — and the subscription that would do it is already paid for. Kew is not the place for this; the GRO and the census are.

21 September 2026 — searched, and the answer is the other surname.

No person of this name, in any spelling, is enumerated anywhere in Glamorgan in any census from 1841 to 1921. Defranceschi and De Franceschi return nothing across Great Britain for births 1815–1895; the wildcard *Defrance\ returns no census record at all** in Glamorgan; and the one historic Cardiff hit on *Franceschi\ is a Franceschini**, a different name.

The Cosulich are there the whole time — Swansea 1871, St John Cardiff 1881 and 1891, Barry 1921, Rhyl 1939. «An old Cardiff family» is the bride's family, not the groom's. The reading is on the search register and the household is set out under Cosulich.

What is left is the GRO, and it is an order, not a search: the 1876 marriage certificate (Cardiff Q3 1876, 11A/428) for Emerico's occupation, and the Cardiff birth index of the late 1870s for the daughter who played at the 1896 Exhibition.

Begun 17 September 2026, and the survey of the shelf came first because it changed the target.

Across the whole Vinodol, one marriage register on film reaches the years wanted. Ledenice's Vjenčani stops at 1882; Novi Vinodolski's at 1859; Bribir's at 1858; Grižane has no marriage film at all. Only Crikvenica, Vjenčani 1859–1908, covers 1885–1890 and 1908. If the marriage of Josip Defranceschi × Anna Kalanj is on free film anywhere, it is in that one volume — and if it is not in it, the answer is a letter to the parish office at Ledenice or Novi, not another search.

The film is open and mapped. 272 images, DGS 2084742; every image's ark captured from the waypoint API and held in the session scratchpad; full-resolution pages (about 6000 × 4150) reachable directly, and legible. Year anchors read off the pages themselves: image 61 = 1862 · image 161 = «Godina 1890» · image 200 = 1897, so 1885–1890 falls at about images 138–165 and 1908 at the end of the book.

The honest cost of «end to end». The archive's own standard is that a volume is not marked read until every page has been looked at. A page is a two-page spread of about four marriages, and the browser returns 800 pixels at a time, so a spread needs three or four looks. 272 images is therefore something like nine hundred reads — an afternoon, not a minute, and it is the price of being able to put a green mark on that volume and mean it.

20 September: 17 of 272 read, and three consecutive years are now closed end to end.

1887, 1888 and 1889 are complete — images 151–160, entries 1–19, 1–19 and 1–13, with the priest's own closing line on each: «U Crikvenici dne 31 prosinca 1889. Alexander Car, župnik.» Not one Defranceschi in any of them, in the couple columns or the parents' and witnesses' columns.

Why those three years and not others. Anton Rudolf was born at Klenovica on 20 January 1890, so his parents Josip Defranceschi and Anna Kalanj married before that; Ledenice marriages stop in 1882, which leaves Crikvenica as the only Vinodol marriage register on film covering the window. 1887–1889 is the heart of it, and they are not in it. That does not close the question — 1883–1886 and the first weeks of 1890 are still unread, and they may have married in Klenovica's or Novi's own parish rather than the bride's — but it removes the likeliest three years.

One gap to go back for: 1888 entries 16 and 17 fall between images 157 and 158 and have not been seen. The year is otherwise complete.

The scratchpad was wiped between sessions and had to be rebuilt — all 272 arks re-derived from the waypoint and verified against the previous session's record (image 156 must be 3QS7-899X-R4S, and it is). METHOD.md now carries the rebuild recipe so the next wipe costs minutes.

The reading method changed and is much better. The old auto-crop locked onto the film frame rather than the page, because this film's grey base is as bright as the paper. It is replaced by a fixed left/right split — each half of the frame drawn into 800×1100 — which doubles the effective resolution and made the hands legible. Two screenshots per image; about 255 images to go.

And one question for David, which is a scope question rather than a reading one. The witness column of 1888 no. 18 carries «Frančić». The declared in-scope list at /name/#scope includes «Franceschich», which is the Italianised spelling of exactly that Croatian patronymic of Franjo. Either Frančić is in scope and this film has hits, or Franceschich should not be on the in-scope list. It cannot be both. Flagged, not decided.

20 September, second sitting: 1883 is complete, and a year label had to be corrected before it did any harm.

The correction first. Image 132 was logged an hour earlier as «Godina 1885». Read at full magnification it is unmistakably «Godine 1883» — the same flourished hand that had already turned 1888 into 1877 at low resolution. The digit was re-read, not guessed, and everything then fell into place: 1882 closes on image 131 with «Crikvenica dne 31 prosinca 1882», so the next year must be 1883. 1883–1886 is images 132–150, which is four years across nineteen images — the same four or five images a year as the rest of the film.

1883 is now read end to end: images 132–136, entries 1 to 23, and no Defranceschi. The year closes with blank ruled rows after entry 23, which is how this register ends a year. The names are the same closed village as ever — Car, Kalnić, Škiljan, Kristofor, Vračić, Bernjac, Škomerža, Kulentić, Matejčić, Manestar — with two outsiders worth noting: no. 19, Josip Car, a clerk at the royal district tax office at Karlovac, and no. 21, Ivan Barac of Grižane marrying Ana Car of Crikvenica.

1881 and 1882 were read on the way past (images 124–131) and are also clean, which was not asked for but costs nothing to keep.

Read so far: 31 images of 272, and 1883, 1887, 1888 and 1889 are complete years. Still to do for David's window: 1884, 1885 and 1886 — images 137 to 150.

Method note: the crop is now a magnified window on the left page rather than a whole-frame view, because the couple columns are what carry the surname. The parents' and witnesses' columns on the right page are a second pass, still owed for 1883.

1883 to 1889 is now read end to end — seven consecutive years, 144 marriages, and not one Defranceschi.

Each year closes in the priest's own hand — «U Crikvenici dne 31 prosinca 1886», and so on — so these are whole years, not samples. 1881 and 1882 were read on the way in and are clean too.

What that settles. Anton Rudolf was born at Klenovica on 20 January 1890; his parents Josip Defranceschi and Anna Kalanj must therefore have married by 1889. Ledenice marriages stop in 1882. Crikvenica was the one Vinodol marriage register on film that covered the gap, and the seven years that matter are now read and they are not in it.

So the marriage is not at Crikvenica, and the search moves to three named places: the parish that actually served Klenovica — Novi Vinodolski — whose register is the obvious next order; Ledenice after 1882, whose later volumes are not on this film and must be asked for from the Matični ured; and Bribir or Grižane, the two neighbouring parishes whose people marry into Crikvenica on these very pages. This is a redirection built on a complete reading, not on a sample.

The village these pages describe is worth recording for its own sake. Ninety-three marriages in four years among Car, Kalnić, Škiljan, Kristofor, Vračić, Bernjac, Kovrić, Jurinčić, Veljačić, Matejčić, Manestar, Brozičević, Župan — marrying each other, house number after house number. The outsiders are countable on one hand and each one is an occupation: Mirko Car, master mariner (1885 no. 24); Adalberto Radil, a baker (1885 no. 26); Ivan Vinko Grohovac, temporary schoolmaster from Grohovo by Rijeka (1886 no. 4); Eduardo Janić of Bakar (1886 no. 7); Josip Car, clerk at the royal tax office at Karlovac (1883 no. 19); and Marija-Eva Pilz of Gospić (1886 no. 10), the only German name in four years. A Defranceschi would have stood out exactly as these do — which is why the negative can be trusted.

Read so far: 45 of 272 images. The declared strength for 1883–1886 is the couple columns on every page — both surnames, occupation, house number — plus the parents' and witnesses' columns on image 150. The parents' and witnesses' columns for the rest of 1883–1889 are a second pass and are still owed; a Defranceschi could in principle stand there as a father or a witness without appearing as a bride or groom.

Second pass done. 1883–1889 is now read on BOTH pages — every column of every entry — and the name is not there.

The right-hand page of this register carries Parentum Nomen (both sets of parents, with the mothers' maiden names), Testium Nomen (two witnesses with their fathers' names and trades), the celebrant, the banns and dispensations, and the Observationes. Twenty-five right pages were read to finish images 132–157; 150 and 158–160 had been done in the first pass. 144 marriages, seven years, four columns of names each, and no Defranceschi anywhere in them — not as a groom, not as a bride, not as a father, not as a mother, not as a witness.

That is the strong form of the negative. A man can marry out of a parish and never appear in its register; he cannot easily be a father or a witness there either without leaving a trace. Josip Defranceschi is absent from Crikvenica in all four roles across the whole window in which his son was conceived and born. The marriage, and the family, were somewhere else.

Things the right pages gave that the left ones could not. The parish had three priests, not one: Aleksander Car signs almost everything, but Vč. g. Mate Škiljan, umirovljeni župnik — a retired priest — takes services on images 148, 149, 153 and 155, and Lovro Zenac, kapelan Grižki, comes over from Grižane for 1885 no. 24. Two marriages needed military permission — «dozvolom vojničke oblasti», from Rijeka (1884 no. 16 and 1886 no. 6), which means a serving soldier. And the witness column is where the village's professional class shows up: Josip Šanvić, obćinski bilježnik (parish notary), Mate Car, obćinski pazitelj and kr. lučki pazitelj (harbour watchman), a trgovac, and on 1888 no. 12 Narcisio Kamin, professor of navigation — a nautical school teacher standing witness at a Crikvenica wedding.

The scope flag is now on two pages, not one. Witnesses spelled «Franić» / «Frančić» appear at 1888 no. 6 and 1888 no. 18. The declared in-scope list still contains «Franceschich», which is the Italianised form of that same Croatian patronymic of Franjo. Ruled on 20 September 2026: «frančić is adjacent, take franceschich off the list». So they are not hits, and the seven-year negative for 1883–1889 stands unqualified. Franceschich has been moved to the adjacent forms and taken off the drift chart — correction 110. Twelve spellings are the name; six are adjacent.

Into the second half: 1890 and most of 1891 read, both pages, on the way to 1908.

1890 is complete — images 161–163, eleven marriages, closed «U Crikvenici 31 prosinca 1890». 1891 is read to entry 23 across images 163–167. No Defranceschi in either year, in any of the four name columns.

The village is beginning to leak, and the register shows it. Where 1883–1886 had five outsiders in four years, 1890–91 already has a bride from Senj with the župnik of Senj himself standing witness (1890 no. 9), a groom from Vrbnik on Krk married by permission of the Vrbnik parish office (1890 no. 11), a man from Sv. Križ Začretje in the Zagorje who is a čuvar na željeznici u Rieci — a railway guard at Rijeka (1891 no. 14) — and brides and grooms out of Grižane and Sv. Jelena. The railway and the steamer are pulling strangers into a parish that married itself for centuries, and that is the context in which a Defranceschi could arrive.

And the bishop came. Image 165 carries a seal and a signature across the foot of the page: «Vidjeno na kanoničnoj vizitaciji dne 20. svibnja 1891 — Juraj Posilović, biskup» — the Bishop of Senj-Modruš inspecting the register in person. It is the second visitation mark on this film; there is another, unsigned, in 1881.

53 of 272 images read. 1883–1891 is now continuous and complete except the last stretch of 1891. Still to do: 1892–1908 (images 168–272) and 1859–1882 (images 4–123).

1891 and 1892 complete; 1893 begun. 59 of 272 images.

1891 closes at 24 marriages (images 163–168) and 1892 at 23 (images 168–172), each signed off by Aleksandar Car on 31 December. 1893 is open at entry 5. No Defranceschi in any column of either year. 1883 to 1892 is now ten unbroken years.

The outsiders have stopped being a trickle. In 1892 alone: a joiner from Vrbnik on Krk; Dragutin Gaus, another joiner, married by leave of the royal Hungarian ministry of defence; and — the first of its kind in this book — Marko Pavletić, an innkeeper, marrying Zuzana Rotter of Nádasd in Hungary, the pair of them entered as «necatolik». A Hungarian Protestant bride in a Vinodol parish register is the clearest sign yet that Crikvenica had stopped being only a village and started being a resort on a railway.

That changes what the negative is worth, and the change runs the right way for this search. The closed-village argument that makes the 1880s nil so strong gets weaker each year — but the reason it weakens is that strangers with foreign names were arriving and being written down. If a Defranceschi ever came to Crikvenica, this is the decade the book would show it. So far it does not.

1893 and 1894 complete, 1895 read to entry 20. 71 of 272 images. Twelve unbroken years, 1883–1894, and still nothing.

1893 closes at 23 marriages, 1894 at 18. No Defranceschi in any column of either.

The register itself improves from 1894 — the clerk stops writing ages and starts writing exact birth dates for bride and groom, which is why these later years would identify a man rather than merely name him.

And the arrivals keep widening. A pharmacist from Kraljevica (1893 no. 17); a man from Bag — Karlobag (1893 no. 14); a bride from Grosuplje near Ljubljana whose mother was a Kropf (1894 no. 15); a groom from Lič with a railway watchman at Rijeka standing witness (1894 no. 16); a wedding performed at Kraljevica by its own chaplain (1895 no. 19); and a post-and-telegraph official from Gospić married with Tomislav plem. Budisavljević — a nobleman — as witness (1895 no. 20).

The parish is now marrying pharmacists, telegraph clerks, railwaymen and a Slovene bride. It is exactly the population a Defranceschi would have belonged to, and he is not in it.

1896 and 1897 complete, 1898 read to entry 24. 89 of 272 images. Fifteen unbroken years, 1883–1897.

1896 closes at 22 marriages, 1897 at 26. No Defranceschi in any column of either.

One old error corrected on the way past. A previous session logged image 200 as «1897 nos 1–2». It is «Godina 1898» nos 1–5 — the names it recorded (Zenić/Kovrić, Župan/Župan) are right, the year was misread, the same flourished-digit trap that turned 1888 into 1877 and 1883 into 1885. The log now carries the correction.

And the town has finished changing. 1898 alone gives a hotel supervisor, a professor, a bank director and the municipal treasurer standing as witnesses, a telegraph worker at Rijeka, a merchant from Rijeka named Benvenuto Vekil — an Italian forename at last — and a page on which nearly every groom is a zidar, a mason. Crikvenica in the 1890s is a building site with a hotel on it, and its register is full of masons, carpenters, tinsmiths and hotel staff drawn in from Krk, Ljubljana, Gmunden, Karlovac and Gospić.

That is the strongest form the negative can take. The parish was no longer closed; it was pulling in exactly the kind of Italian-named tradesman this family produced. Fifteen years of it, and the name does not appear once.

1898 and 1899 complete, 1900 read to entry 20. 100 of 272 images — and the film is now more than a third read.

1898 closes at 24 marriages, 1899 at 32. Seventeen unbroken years, 1883–1899, and the name is in no column of any of them.

A second bishop signs the book. Image 210: «Vidio na kanoničkoj vizitaciji u Crikvenici dne 20. svibnja 1899. + Antun» — Antun Maurović, Posilović's successor at Senj-Modruš. That is two episcopal visitations caught in this reading, 1891 and 1899.

By 1899 the register has become a list of masons. Page after page: zidar, zidar, zidar. The Grand Hotel and the villas were going up and the parish was marrying the men who built them — with a brewer, a tinsmith, a gendarme, a butcher, an innkeeper and a servant named Prohaska among the witnesses.

Which is the whole point of reading it. Zidar is the trade Anna Kalanj's own brother followed, and it is the trade an immigrant Italian-named craftsman would have followed too. Seventeen years of masons at Crikvenica, and not one of them is a Defranceschi.

1900, 1901, 1902 and 1903 complete. 113 of 272 images — twenty-one unbroken years, 1883–1903.

1900 closes at 24 marriages, 1901 at 21, 1902 at 26, 1903 at 28. No Defranceschi in any of the four name columns of any of them.

1902 has a five-month hole in it that is not a gap in the reading. The book runs from 19 February straight to 22 October — no marriage at all between them. That is Lent and then the building season: in a parish of masons and mariners, nobody marries while the work is on.

The catchment is now continental. These four years bring in a groom from Kowalówka in Galicia, Josip Elia Schmidt, son of an officijal, marrying a Car of Crikvenica (1902 no. 25); a carpenter from Gödöllő and a railwayman from Darázsfalva, both in Hungary; a man from Krupa in Bosnia; a groom from Medjimurje, parish Prelog; one from Ovba in Ungarija; and a Crikvenica joiner whose domicile is entered as «u Poli» — Pula — married by leave of the Pola parish office. Three entries of 1903 are in a second hand with a different celebrant and the religion column marked «P. Kt.», a non-Catholic rite written into the Catholic book.

One line is worth recording even though it is not our name. 1901 no. 10, image 218: the groom's origin column reads «Landrovac» with «Italij[a]» struck through — the only mention of Italy anywhere in this book so far. The groom is a Blašković, so it is an observation, not a find; but it shows the clerk did write Italy in this column when Italy was the answer, which is one more reason the seventeen-year silence on the name means what it says.

Still to read: 1904–1908 (images 233–272) and 1859–1882 (images 4–123).

1904, 1905, 1906 and 1907 complete. 144 of 272 images — twenty-five unbroken years, 1883–1907, and one page of 1908 to go.

1904 closes at 26 marriages, 1905 at 36 («Crikvenica dne 30 prosinca 1905. Car, župnik»), 1906 at 45 — the busiest year in the book — and 1907 at 31. No Defranceschi in any of the four name columns of any of them.

These four years are where the foreign names finally arrive in force, and they are worth listing because each one is the kind of entry a Defranceschi would have made:

• Josip Wolf, a locksmith from Novi Sad in the Bačka, living at Crikvenica, marrying a Vračić (1907 no. 30);
• Berti Königsberger, son of Herman Königsberger and a mother née Feiner, marrying a Car (1906 no. 36) — the first German-Jewish surname in the register;
• Ferdinando Sandor Mihely Kolomarić, a Hungarian railwayman routed through Zagreb (1906 no. 32);
• Franjo Ivan Pagliaga of Bakar, son of Anton Pagliaga, mariner, and Vittoria (1907 no. 3) — an Italian surname on an Italian mother, read at magnification to be sure;
• Milan Caran of Korenica, domiciled at Pola, entered as istočno-pravoslavne vjeroispovjedi — an Orthodox groom married to a Catholic bride under naval permission from Pola (1905 no. 30);
• a Dalmatian ex-soldier at Pola (1904 no. 1) and a joiner from Pag (1906 no. 35);
• grooms from Gödöllő, Darázsfalva, Debrecen, Nagy Torda, Nedelišće, Vrbovsko, Novi Vinodolski, Medjimurje and Matarija;
• and Italianate forenames on local surnames — Ninfa Brnjac, Libera Matejčić, Emigdio Milan Brnjac, Matilda Car, Kalista Tracić, Roberto Paulin, Rudolfo Car, Eduardo Brnjac.

That is the test this reading was for. The parish is now marrying a Jewish-named clerk, a Serbian Orthodox sailor, a Hungarian railwayman, an Italian mariner's son from Bakar and a locksmith from Novi Sad — and writing every one of them down, with birthplace, father's name and trade. An Italian-named mason called Defranceschi is precisely what this book records when it happens. Twenty-five years, and it never happens.

One caution logged against myself: 1907 no. 20 looked at first like «iz Italije». Magnified, it is «i Kalista» — the bride's forename. The correction is in the reading log; nothing was published on the misreading.

Still to read: the last of 1908 (images 262–272) and 1859–1882 (images 4–123).

The second half of the film is finished. 1883 to 1908 is read end to end, both pages, and the name is not in it.

1908 closes at 33 marriages (images 262–268), and image 272 is the film's own closing target: «ŽUPA CRIKVENICA · HRVATSKA CROATIA · MATIČNA KNJIGA VJENČANIH 1859–1908 · ITEM 1 · project CRO 00008 · roll 28». There is nothing after it.

Twenty-six consecutive years — 1883 to 1908 — every entry, every column. Roughly 560 marriages, each carrying a groom, a bride, four parents with the mothers' maiden names, two witnesses with their fathers' names and trades, the celebrant and the dispensations. No Defranceschi stands in any of those roles, in any year.

The last three pages are not entries at all, and they are worth having. Images 269–271 are blank ruled pages with loose certificates pasted onto them — the parish's own working papers, kept in the back of the book:

• a freedom-to-marry certificate of 16 November 1868 for a Tomas Škrolin-Lončar of Crikvenica, signed Josip Medvedić, crikvenički vikar;
• another of 22 November 1869 in the same hand;
• a medical clearance for a Mihovil Car, «Crikvenica 20/4 66»;
• and, dated 22 April 1866, a note about Ludoviška Car of Crikvenica signed «Katarina Šuvčić, babica» — the village midwife, in her own handwriting. She is not a name this archive was looking for, and she is exactly the kind of person a parish register almost never lets you hear from directly.

And one entry that shows how far the parish would reach when it had to. 1908 no. 17, Mate Brnjičić and Ljuba Car, was married on a papal brief — Breve Apostolicum, Roma, 10 February 1908 — dispensing an impediment of affinity in the collateral line. Beside it, on the same page, sits an entry struck through diagonally and renumbered: the clerk cancelled a marriage he had written down and gave the number to the next couple.

What is left. The first half — images 4 to 123, the years 1859 to 1882. Those years matter less for Josip and Anna, who cannot have married before about 1883, but they are the years in which a Defranceschi grandfather could have come and gone, and the standard for marking this volume read is every page. The reading continues at image 4.

The first half is open, and the book turns out to be in Latin.

Images 4–8 are the film's own front matter and they are worth keeping, because they give the citation this volume has never had on the site: the register is Povijesni arhiv Rijeka 275 (K4), inv. 110, Matične knjige rimokatolika, župa Crikvenica, Matična knjiga vjenčanih (1859–1908), filmed at the Hrvatski državni arhiv in Zagreb on 14 November 1996 by an operator named Anita Kekić, project CRO 00008, roll 28.

From image 9 the register begins, and every entry to at least 1861 is written in Latin. Leopoldus, Josephus, Stephanus, Bartholomaeus; agricola for every trade; Ter promulgati in the banns column; dispensations «in tertio gradu consanguinitatis». That changes what the eye is looking for. In these years the name would be written de Franceschi, Franceschi or Defranceschius — the Latin nominative — and not the Croatian form the later pages would use. The reading is being done with that in mind.

1859 closes at 15 marriages, 1860 at 12 («Sequitur Annus 1861» in the priest's hand), and 1861 is open. Nothing so far. Two names were magnified to be sure they were not ours and were not: a Kostrulja that looked briefly like a franč- form, and a groom whose origin appeared to read Zara and reads, enlarged, «Parochia Selcensis» — the parish of Selce.

FamilySearch has flagged these pages «WATER DAMAGE / OŠTEĆENO OD VLAGE», and the flag is honest: the ink is pale and the margins are cockled. They are still legible.

1861 and 1862 complete; 1863 read to entry 12. Images 4–30 done.

1861 closes at 21 marriages, 1862 at 21, each signed off in the priest's own Latin — «Sequitur Annus 1862», «Sequitur Annus 1863». No Defranceschi in any column of either.

Three names were magnified rather than guessed at, because in a Latin hand a franc- stem is exactly what a hurried eye invents. 1861 no. 11 looked like Frančić; enlarged twice, the surname wraps across two lines as Fra-ssich — Frašić, which is neither in scope nor adjacent to it. The other two resolved the same way.

What these pages give besides the negative. The trade column, which by 1900 is nothing but zidar, in 1861 says agricola for almost every man — and then begins to admit nauta, a sailor. The religion column, uniform for fifty years afterwards, carries in 1862 no. 10 a second line reading «Cath. Graeca» beside Rom. Cath. — the only non-Latin-rite marking in the book so far. Brides and grooms come in from Selce, Grižane, Zagorje, Sv. Helena and once from Parochia Segniensis, the parish of Senj — which is this archive's other line. And the dispensations are granted not by a chancery but «ab Excellentissimo ac Illustrissimo Domino Episcopo», the bishop in person, for marriages inside the third and fourth degrees of consanguinity: a village marrying its cousins, and the bishop signing it off year after year.

1863, 1864 and 1865 complete; 1866 read to entry 8. Images 4–44 done.

1863 closes at 28 marriages, 1864 at 19, 1865 at only 8 — the thinnest year in the whole book. Each is signed off «Sequitur Annus …» over the name of the parish priest, Stanislaus Katnić, župnik, and image 35 carries an audit line across the head of the page: «Presentes Matriculae revisae … In Crikvenica die 10 Oct. 1863 — Venceslaus Car, curatus.» No Defranceschi in any column of any of them.

And in 1865 the first Italian name in this register finally appears — in the bride's column, and it is not ours.

1865 no. 7, read at magnification because the hand is small: «D. Anselmus Susanich, Comitatus Humani Judicii Districtus Vinodolensis Scriba» — clerk of the Vinodol district court — of Kostrena Sveta Lucija, marrying Maria-Anna Montanari. One of the witnesses is Antonius Sokol, negotiator, a merchant. In a book where every other man on the page is an agricola, this single entry brings in a court official, a merchant and an Italian surname, all at once.

That is the shape of the evidence, stated plainly. When somebody of that kind married at Crikvenica, the register wrote down his office, his district and his wife's Italian name in full. It had the vocabulary and it used it. It never had occasion to write ours.

Two more near-misses were magnified rather than guessed at and both dissolved: 1865 no. 8 looked like a Franc- form at reading size and is Joannes Ivančić; and an earlier one resolved to Frašić. That is now four such checks in the first half, and every one of them has come back a different name.

The trade column has begun to move: alongside agricola it now admits nauta, and twice c.r. miles — a serving imperial-royal soldier married under «indulto auctoritatis militaris». Witnesses come from de Schola Selcensi, the Selce schoolhouse. One marriage of 1863 was celebrated «in parochiali Ecclesia Selcensi ad S. Jacobum», in the church of St James at Selce.

1866 complete; 1867 read to entry 37. Images 4–58 done — a fifth of the first half.

1866 closes at 20 marriages. 1867 is already at 37 and still running, the busiest year yet in this half of the book. No Defranceschi in any column of either.

The parish is going to sea. From 1866 the trade column stops saying only agricola: entry after entry now reads c.r. nauta — an imperial-royal sailor — or c.r. miles, and their marriages carry «cum indulto c.r. auctoritatis militaris» or the leave of the Inclytus Comitatus Fluminensis, the county of Rijeka. In 1867 alone, seven grooms are serving men. And 1866 no. 12 brings the first ship's officer into the book: «D. Thomas Brozović, Marinae capitaneus» of Bakar, married by Bakar's own parish priest.

One page is not a record at all. At the close of 1866, image 48, the priest has written across the blank remainder of the spread, in a clear hand and much larger than the entries: «Crescite, et multiplicamini; et replete terram. — Gen.» — Increase and multiply, and fill the earth. Then, beneath it, «Sequitur Annus 1867».

A fifth magnified check: 1867 no. 10 read as Franich at reading size and is Dračić. Five checks in the first half, five different names, and none of them ours.

1867, 1868 and 1869 complete; 1870 begun. Images 4–68 done.

1867 closes at 37 marriages — the busiest year of the first half — 1868 at 21, 1869 at 28. No Defranceschi in any column of any of them.

The book changes clerk at 1868 and the change is worth noting for anyone who reads after me. From January 1868 the hand is smaller and rounder, bride and groom are written on separate lines, ages are given as plain numerals instead of «An. 24», and the religion column is spelled out in full — Romani Catholici. Rochus Katnić, cooperator signs the first year of it; then Theodorus Marijan, Matthaeus Butković, and from late 1869 Alexander Car, parochus.

The second Italian surname in the book arrives in 1868, and like the first it belongs to somebody passing through. 1868 no. 16, magnified to be sure of it: Antonius Benedetti of Fiume, mercator, marrying Maria Pelc of Hinje in Carniola. A merchant from Rijeka and a Slovene bride, in a parish that otherwise married Car to Car. Again: the register knew how to write an Italian name. It wrote this one.

The Genesis line is not a one-off. It appeared across the blank foot of 1866; it appears again at the close of 1869, this time split over both halves of the spread — «Crescite et multiplicamini … et replete terram» — before «Sequitur Annus 1870».

And the sea keeps taking the men. Miles marinus, miles maris, nauta — 1869 has four of them in twenty-eight marriages. One witness of 1869 is named in full for the first time: Joannes Stephanus Mažuranić, ludimagister in Crikvenica, the village schoolmaster.

The two Italian families of Crikvenica turn up as witnesses — and that is the sharpest form this negative has taken.

1870 complete at 25 marriages; 1871 read to entry 11. Images 4–76 done.

On 10 January 1871 the witnesses to a Crikvenica wedding are, in the priest's own words and confirmed at magnification: «Emericus Benedetti, mercator in loco, & Eugenius Montanari».

Both names have been met before in this reading, and both arrived from outside:

• Montanari first appears in 1865, when a Maria-Anna Montanari marries the clerk of the Vinodol district court;
• Benedetti in 1868, when Antonius Benedetti of Fiume, a merchant, marries a Slovene bride from Carniola.

By 1871 they are no longer passing through. Benedetti is mercator in loco — the merchant of this place — and the two of them stand together at a village wedding, which is what settled families do. The register names them, gives their trade, and states their residence.

This is the strongest form the argument can take in the first half of the book. Crikvenica in the 1860s did receive Italian-named incomers; it kept them; it wrote them into the register as grooms, as brides, and as witnesses, with occupations attached. The parish had both the vocabulary and the occasion. Across 1859–1871 it never once had cause to write Defranceschi.

Also on these pages: 1870 no. 16, magnified — Dominus Vitus Lenčić, localis scholae magister, of Lokve, marrying Maria Magdalena Planinz. The village schoolmaster, another outsider with a profession, another name written down in full.

A third bishop signs the book — and this one is the earliest.

Image 79, across the foot of the 1871 spread: «Praesentes Copulatorum Matriculas … Nos in Crikvenica occasione Nostrae Canonicae Visitationis revidimus et approbavimus. Sic in Crikvenica die 13ª Octobris 1871.» — Venceslaus Soić, Episcopus Segniensis et Modrusiensis.

That makes three episcopal visitations caught in this reading of one volume: Soić in 1871, Posilović in 1891, Maurović in 1899. Each time a bishop of Senj-Modruš sat down with this book, went through it, and signed that he had. Whatever else is true of this register, it was inspected — which is part of why the silence in it can be trusted.

Images 4–80 done; 1871 read to entry 25. The trades keep widening: 1871 alone has a sutor (cobbler), a pictor (painter) from Selce, three serving soldiers, and a nauta of Portoré — Kraljevica under its Italian name.

A sixth magnified check: 1871 no. 15 read as a possible Franc- form and is Matthaeus Franović, sutor. Six checks in the first half, six different names. And 1871 no. 14, first read as «Istria», is «Sponsus Portoregensis».

1872 and most of 1873 read. Images 4–86 done. 1872 closes at only 10 marriages — the thinnest year since 1865.

Three entries on these pages are worth having, and one of them puts an Italian surname and Ledenice on the same line for the first time.

1873 no. 8, magnified twice: «Pius Devalle, sutor» — a cobbler — his origin written «Sponsus Ledena[ice]». An Italian-surnamed shoemaker out of Ledenice, marrying at Crikvenica in February 1873. He is not ours, and nothing here suggests he is. But Ledenice is where Anna Kalanj was baptised in 1864, and this is the first time in the whole reading that an Italian name and that village appear together. It is logged for what it is: a fact about the place, not a link.

1873 no. 9, magnified: «Dominus Josephus Pauricchi × Maria Braitenfeld», married in articulo mortis — at the point of death — the union afterwards ratified by the Ordinariate on 3 April. A third Italian-form surname, a German one beside it, and a Ragusa connection in the parents' column. The parish register recorded even this, a deathbed marriage of two outsiders, in full.

And Benedetti is still here. He witnesses 1872 no. 7 and again 1873 no. 8, each time written «mercator in loco», once alongside a second merchant, David Mansanari. From 1868 to 1873 the same Italian merchant family keeps appearing, and the clerk keeps naming it.

Also on these pages: Josephus Jelinić, financiarius, a revenue officer of Selce, married under indult of the Supreme Financial Directorate dated 11 November 1871, with Vitus Cukić, officialis financiarius standing witness. The Austrian state is now inside this village register too.

1873 and 1874 complete; 1875 begun. Images 4–94 done — the first half is two-thirds read.

1873 closes at 19 marriages, 1874 at 23. No Defranceschi in any column of either.

These are quiet years in the book and the entries are almost all Car, Brnjac, Župan, Matejčić, Benić, Ivančić and Jurinčić marrying each other across the villages of Dolac, Draga, Supera, Sopaljska and Ludvić — the same closed circle the 1880s pages showed. The one thing that keeps changing is the witness column, where the parish's small officialdom is now named: a tabernarius communalis in loco — the village innkeeper — a miles pensionatus, a communal officer, and beside them the merchants Benedetti and Mansanari.

Remaining: images 95–123, the years 1875 to 1882.

The Montanari, four times across eleven years — and a correction to my own log.

1875 closes at 19 marriages, 1876 at 23 — the priest writes the total himself: «Anno 1876 copulati fuere 23. Sequitur annus 1877.» No Defranceschi in either. Images 4–104 done.

1876 no. 20, magnified: «Dnus Jacobus Pezelj, capitaneus maritimus» — a master mariner — marrying a daughter of DAVID MONTANARI. The parents' column gives it plainly: «Dftus Laurentius Pezelj & Maria nata Zelenčić — David Monta[nari] & …»

That is the fourth appearance of this one Italian family in the volume:

A correction to my own reading, made before it did any harm. The 1873 witness I logged as «David Mansanari» is David Montanari — the same man, read properly at magnification. The scratchpad log carries the correction.

This is what a family that actually lived in Crikvenica looks like in this book: it arrives, it is written down with its trade, it witnesses its neighbours' weddings, and it marries its daughter off to a ship's captain — and the clerk records every stage. Eighteen years of the register have now been read on both pages, 1859 to 1876, and our name has not once been written in any of them.

Remaining: images 105–123, the years 1877 to 1882.

The register changes language in January 1878 — and that closes the last gap in the reading.

1877 closes at 16 marriages, signed off «Sequitur annus 1878. Crikvenica 31 decembris 1877. Alexander Car». No Defranceschi. Images 4–110 done; thirteen images left.

At the head of 1878 the book stops being Latin and becomes Croatian. «Ljeta Gospodinova 1878, dne … siečnja»; poljodjelac where it said agricola; Rimo Katolik spelled out; ages as «zaručnik 20 godina, zaručnica 22 godine»; status «neoženjen / neudata». Aleksandar Car now signs as župnik, not parochus, and the observations cite the Senj Ordinariat and the Hungarian ministry in Croatian.

Why this matters for the search, and not only as a curiosity. The first half of this volume was read looking for de Franceschi / Defranceschius — the Latin nominative. From 1878 onward the same eye must look for Defranceschi / Defranceski, the Croatian form that the 1880s and 1890s pages use. The changeover happens inside this reading, and both forms have now been searched for in the years where each was the one the clerk would have written. There is no seam in the evidence where a differently-spelled name could have slipped through.

And in 1878 the trade that ends this book begins. 1878 no. 7: «Mate Kristofor, zidar» — the first mason in the register. By 1899 nearly every groom on the page is a zidar; here is the first of them.

Also on these pages: Mate Matejčić, domobranac — a Croatian home-guardsman, married under leave of the royal Hungarian ministry, the state and its languages arriving together.

The volume is read. All 272 images, both pages, 1859 to 1908 — and the name is not in it.

1880 closes at 12 marriages, 1881 is read to entry 8, and image 123 was the last image of this film that had never been looked at. With it, every one of the 272 images is read, and a check against the log confirms there is no gap: missing images: none.

Fifty years of one parish's marriages, on both pages of every entry. Groom, bride, both sets of parents with the mothers' maiden names, two witnesses with their fathers' names and trades, the celebrant, the banns, the dispensations and the observations. Not one Defranceschi, in any spelling, in any of those roles, in any year.

And the one debt this reading declared against itself is now paid. Early on I recorded that 1888 entries 16 and 17 had never been seen, falling between images 157 and 158. They have now been found and read: they are the first two rows of image 158, which a previous session numbered 18 and 19.

• 1888 no. 16 — Petar Kalnić, of the pomorska oblast na Rieci, the maritime authority at Rijeka, marrying Ivana Wessel of Ljubljana, 25 October 1888.
• 1888 no. 17 — Mikula Car and Marija Car, radnici, 21 November 1888.

Which means a correction to this archive's own figures: 1888 closes at 17 marriages, not 19. The names in that year were right; the count was two too many. It is recorded here rather than quietly fixed, because the point of reading a volume end to end is that the arithmetic can be checked.

What fifty years of this register actually shows

The parish was not closed to strangers, and it wrote every one of them down. Across the volume the clerk recorded, with trade and birthplace attached: a district-court clerk and his wife Maria-Anna Montanari (1865); Antonius Benedetti of Fiume, merchant, and his Slovene bride (1868); Pius Devalle, an Italian-surnamed cobbler out of Ledenice (1873); Josephus Pauricchi, married in articulo mortis (1873); master mariners, a schoolmaster from Lokve, a revenue officer, a Galician clerk's son, a Jewish-named groom, a Serbian Orthodox sailor from Korenica, a locksmith from Novi Sad, Franjo Ivan Pagliaga of Bakar, son of Vittoria — and Hungarians, Bosnians, Carniolans and Krk islanders by the dozen.

Two Italian families settled and stayed. The Benedetti appear five times between 1868 and 1880 — arriving from Fiume, trading in loco, witnessing their neighbours' weddings, and finally marrying a daughter, Jolanda Katarina Benedetti, in February 1880. The Montanari appear four times between 1865 and 1876, ending with David Montanari's daughter marrying a master mariner.

That is the shape of the negative, and it is as strong as a negative gets. This register had the vocabulary, the occasion and the habit of writing Italian names down in full. It did so for two families over two decades. It never once had cause to write ours.

Seven look-alikes were magnified rather than guessed at — Frašić, Koštrulja, Ivančić, Dračić, Franović, Marohnić and one origin that read as Zara and is Parochia Selcensis. Every one came back a different name. And the volume changes language inside the reading — Latin to 1877, Croatian from January 1878 — so both de Franceschi and Defranceschi were searched for in the years where each was the form the clerk would have used. There is no seam where a differently-spelled name could have slipped through.

What this closes, and what it does not

It closes Crikvenica. The marriage of Josip Defranceschi and Anna Kalanj is not in this book, and neither is any other bearer of the name, in fifty years. That is not a sample; it is the whole volume.

It does not close the question. Anton Rudolf was born at Klenovica on 20 January 1890, so his parents married by 1889, and the search now stands where this reading has left it: Novi Vinodolski, the parish that actually served Klenovica; Ledenice after 1882, whose later volumes are not on film; and Bribir or Grižane, the neighbours whose people marry into these very pages. Those are letters to write and registers to order, not films to read.

Item 130 is done.

Vienna answered the 14 September enquiry by quoting for the sailors and declining the soldiers, and the decline is the more useful half. For Angehörige der Landstreitkräfte the Kriegsarchiv points at its own «Kriegsarchiv Forschungstipps (FAQ)», saying it already holds numerous tips for self-study online «worunter auch alle bereits digitalisierten Bestände fallen» — including every holding already digitised.

Four men of the name are waiting in that half. Theodor Defranceschi, k. k. Landwehr, Fähnrich (nichtaktiv), 27th Regiment, seniority 1 January 1907. Alois Defranceschi, k. u. k. Heer infantry cadet, in the rank lists 1916–1918. Eduard Defranceschi, k. k. line infantry, cadet 1859 and Unterleutnant 2. Klasse 1863 — and whether that is one man or two is still open. And Dr Cäsar Peschle, Landwehr medical officer from 1 August 1895.

What is wanted from each is the same single fact: the Geburtsort. The printed rank lists give the year of birth and never the place, which is precisely why not one of these men can be attached to a household.

Free, online, and never once opened. The paid navy search on Julius and Valentin is a separate decision and costs €45 whether it finds anything or not.

Done, 20 September, and it went further than the four soldiers. The digitised holdings Vienna pointed at turned out to include the whole ÖStA catalogue as a free name search — archivinformationssystem.at/feldsuche.aspx, no account. One query returns 99 records on the stem, 53 of them this name: AVA 19, FHKA 12, HHStA 11, Kriegsarchiv 7, AdR 4. The full inventory is on the research log.

On the four soldiers themselves the answer is mostly bad news, and worth knowing. Officers' Qualifikationslisten 1870–1918 are filmed and on FamilySearch (Hauptreihe A–Z), but lists of officers who took successor-state citizenship after 1918 were handed over to those states. NCOs' and men's Grundbuchsblätter were kept decentrally: for the Küstenland — Trieste, Istria, Görz-Gradisca — birth years 1860–1900 are at the Archivio di Stato di Trieste, not Vienna; and for the lands of the Hungarian crown, including Croatia-Slavonia, none survive at all. Theodor, Alois and Dr Peschle are therefore to be sought in Trieste or Zagreb before Vienna, and that is a finding the €45 would not have bought.

Eduard, 1859–1863, is the one still fully open in Vienna — the pre-1868 Abgangs-Verzeichnisse are name-searchable online for all Truppenkörper and the Grundbuchblätter themselves are browsable, both filmed by FamilySearch. That is the next sitting, and it is free.

Erteilung der Flaggenpatente, Scontrini und Passavanti, nach Namen — a Hofkammer series that registers who was allowed to fly the imperial flag, arranged by name, and the entry reads «… Fiumaner Handelskompagnie 1751-1787 … Franceschi 1759-17.. …».

This archive has an at-sea page built out of masters' certificates and a Cardiff marriage, and it starts in the nineteenth century. A flag patent is the licence a shipowner held. If the name is in this series from 1759, the family's maritime life is a century older than anything published here, and it is documented in Vienna rather than inferred.

The company named beside it matters too. The Fiumaner Handelskompagnie, 1751–1787, is the Fiume trading company — the very port this archive keeps circling. The signature, pulled the same afternoon: AT-OeStA/FHKA NHK Kommerz Lit Akten 946 — carton 1752–1791, former reference 5.377 / r.Nr. 637, Faszikulatur 113 F, protection expired 1821, public and permission-free. And the dates are exact: Franceschi 1759–1770.

The series is the point. Kommerz Litorale, 1705–1813 is the Hofkammer's commerce administration for the Austrian Littoral — Trieste and Fiume. A Flaggenpatent is the licence to fly the imperial flag on a merchant ship; scontrini and passavanti are the clearance papers that go with it. So a Franceschi was licensed to trade under the imperial flag out of the Littoral between 1759 and 1770 — a hundred years before the earliest master's certificate on the at-sea page, and sitting in the same carton as the Fiumaner Handelskompagnie, 1751–1787.

22 September 2026 — checked in the re-run, and it is still one carton and still the only one. AT-OeStA/FHKA NHK Kommerz Lit Akten 946 is the sole flag-patent record of the name in the whole catalogue; the contains re-run that turned up nineteen other files added nothing to this one. It remains an order, not a read.

The arms page rests on a lintel and a stone. The Adelsarchiv holds paper.

AT-OeStA/AVA Adel HAA AR 251.11 and 251.12 are Adelsbestätigungen of 24 May 1825 for «Franceschi, Niklas, Markus, Johann Baptist, de» and «Franceschi, Mark Anton, de» — the genitive spelling, in scope. AR 251.13 confirms Vinzenz Franceschi, 25 June 1830; AR 251.14, Philipp, 1838–39.

And AR 251.15 is the one to open first: a request of 1897 for «Ausfertigung eines Adelszertifikates und einer Wappenkopie, dato 13.05.1825» — a copy of the coat of arms, made from an 1825 original. Behind it sit AR 251.9 (Jakob von Franceschi, Dutch knighthood, 1608), AR 251.10 (Thomas von Franceschi, 1610) and RAA 119.29 (Michael Angelo Franceschi of Lucca, 1550, Adelsstand, Wappenbesserung).

One honest deflation, checked before publishing it. AR 251.15's own catalogue entry reads «Wappen: nein · Siegel: nein» and Extent: fol. 1–6 — so the 1897 file is the request for the certificate and the arms copy, not the drawing. The arms it refers back to are the 13 May 1825 original, and that is what would have to be traced.

A fifth document may matter more than all of them. AT-OeStA/HHStA KA Vertrauliche Akten 56-1-49 — «Nr. 49: Marchese Johann Maria Gravisi, Nicolaus De Franceschi», 1825, fol. 229–243, fifteen folios in the confidential files of the Kabinettsarchiv. The Gravisi were marquises of Capodistria. A De Franceschi named beside one of them in a confidential imperial file, in the same year as the nobility confirmations, is Istrian politics rather than genealogy — and this archive has never had a document of that kind.

Whether any of these is this family is exactly the question. All are public and permission-free.

AT-OeStA/AVA Justiz OJSt JS DS 83.1.14 — «Franceschi, Pietro, Praktikantenstelle in Almissa», 1828–1843. Almissa is the Italian name of Omiš, and the Omiš line is one of this archive's six.

It does not stand alone. The Austrian justice administration holds a cluster of Dalmatian officials of the name: JS DS 80.3.4, Franceschi, Antonio, 1843; JM Präs. A 140.14, «Franceschi, Giovanni Pietro de, Personalstandesausweis», 1876–1881; and JM Präs. A 134.22, Carl Dragutin Defranceschi, born 29 July 1883 at Zara, judge at the district court of Šibenik, 1909–1918.

Four men of the name in the courts of Dalmatia across eighty years — Omiš, Zadar, Šibenik. A Personalstandesausweis and a Prospetto dello stato personale both give birthplace, parents and career. Found 20 September; none of the four is in this archive.

22 September 2026 — the cluster grows and one of them has a birthplace. The ÖStA catalogue, re-run with the «contains» operator instead of the form's default, adds Josef Franceschi, Korporal — «Geburtsort: Almissa, Dalmatien» (AT-OeStA/KA Marine NMA Pers VerEv KL-GagoRangkl 725). A man born at Omiš, in a state record, is new here.

And this roster already carried, from the FamilySearch index, an Antonio Franceschi born 1801 and buried at Almissa in 1877 and another born 1818, buried there in 1888. So the Omiš cluster is now documented from two directions — the parish index and the Austrian service records — without either knowing about the other. No kinship is claimed between them. All are on the roster.

Gallignana is Gračišće, the Istrian parish this family comes out of and whose registers open in 1644. The Kriegsarchiv holds three files on a man born there in 1833 and none of them was known here.

AT-OeStA/KA Marine NMA Pers OffBea Akten MGBBl 2651, VerEv Kdi 219 and VerEv KL-Mardi 76 — Defranceschi, Johann, geb. 1833, Geburtsort Gallignana. The rank moves across the three: Marinediener 4. Klasse on one, 1. Klasse on the other two, so a career is legible from the catalogue alone. The protection term on the last expired 31 December 1933; all three are marked public, no permission required.

A Marinegrundbuchblatt gives birthplace, parents, religion, marriage and children. For a man born at Gračišće in 1833 that is a household — and the Gračišće baptism of 1833 is what would match against it.

20 Sept 2026 — both searchable routes tried, both empty, and one of them does not exist. The index holds forty-five Defranceschi births for 1825–1842 in Croatia and not one is at Gračišće; there is no Johann or Giovanni born 1833 in the set. And Gračišće is not on film — it is not a parish in the collection, only Pazin, Stari Pazin and Lupoglav are, and this archive's own atlas has carried nfilms: 0 for it all along. The sentence that said otherwise has been corrected.

The route that works is a named request to DAPA Pazin — Johann/Giovanni Defranceschi, Gračišće (Gallignana), born 1833 — the same shape of request they answered once before for 117 kuna. That can go in the same envelope as the ÖStA order for MGBBl 2651, and between them a Marinegrundbuchblatt and a baptism would give the man two independent households.

22 September 2026 — the three files disagree about the village, in a useful way. Read at the detail level: VerEv Kdi 219 gives «Gallignano, Istrien» and Marinediener 4. Klasse; MGBBl 2651 gives «Gallignana, Küstenland» and 1. Klasse; VerEv KL-Mardi 76 gives «Gallignana, Istrien, Illyrien» and 1. Klasse. One man, one village, three clerks and three administrative vocabularies — Istria, the Littoral, Illyria — which is the Habsburg reorganisation of this coast showing up inside one sailor's paperwork.

He is now on the roster with all three signatures. The route to his baptism is unchanged: a named request to DAPA Pazin, because Gračišće is not on film.

This archive has six lines and none of them is San Vincenti. Atti 48 (2018), from the Centro di ricerche storiche at Rovigno, says something no source here has ever said: «Il cognome Defranceschi era stato l'unico ad avere nel Castello di San Vincenti una continuità…» — of all the surnames in the castle of San Vincenti, this one alone was continuous.

And it acted as a body. «Guidati da Antonio Defranceschi, nel 1780 avevano ottenuto l'esonero dal partecipare alla difesa armata» — led by Antonio Defranceschi, in 1780 they got themselves excused from armed defence duty. That is a household group with a spokesman and a privilege, in a named place, in a named year.

Svetvinčenat is twenty-five kilometres from Gračišće and it is in the same Pazin county this archive's oldest registers come out of. Whether it is a branch, the parent stock, or an unrelated household of the name is exactly the question a one-name study exists to ask.

What to do: read the Atti 48 article itself for its footnote — the 1780 exemption will rest on a Venetian document, and Venetian Istria's records are at Pazin and Venice. The Svetvinčenat parish registers are the other half.

20 Sept 2026 — the register is filmed, and it goes back to 1568. Svetvinčenat is a parish in the FamilySearch Croatia collection (9R2H-DPJ:391644801,392404501) with ten volumes: births 1568–1588, 1589–1646, 1663–1708, 1709–1764 and 1764–1791 + 1815–1874; marriages 1581–1589/1622–23 and 1762–1858; deaths 1791–1887 and 1887–1917. Unlike Gračišće it needs no letter and no fee. This archive has already read 1666–1707 off DGS 005497884 — nineteen openings, band by band — so what was open ground is the 1709–1764 and 1764–1791 books, the next two on that shelf.

Antonio's 1780 is already bracketed. Births 1764–1791 is 9R2H-PTL:…392405201, 369 images; image 92 is «Anno D.ni 1782, Die 5 Januarij», so at about ten frames a year 1780 is near image 72. Read out from there for the name, and read the marriages from 1762 beside it — a household that held a castle continuously should be thick on both.

Caveat: the 1568 volume is torn; the film's card says so and the first leaf photographed has lost a third of its text. The clean run starts with 1709–1764.

And then it was read. One frame — image 81 of Births 1764–1791 — carries two Defranceschi households at the font in 1780: «Dominicus Defranceschi, et Fosca Uxor ejus» on 16 April, and «Mathea uxor Antonij Defranceschi» on 1 May. The Atti article's Antonio of 1780 is a man on a page. Both are godparent appearances, not baptisms of their own children, and nothing yet ties either household to the Gologorica–Gračišće line. Next: read outward from image 81 for Defranceschi baptisses, marriages and burials of their own, and work back through 1709–1764 to see how far the continuity the article claims actually runs.

And back another fifty-two years. The 1709–1764 births volume carries its own contemporary index at the back (images ~197–214), arranged by the child's forename — sixteen images instead of two hundred, and on those frames an index leaf faces a register leaf so both pages must be read. Its C page gives «Cattarina fig.a di D.o Batta Defranceschi 11 Feb.ro 1728 — c. 66», and carta 66 (image 77, right page) has the entry in full: «Catta fig.a leg.ma di S.r Batta de Fran.chi et di Anzoletta sua Consorte, nata li 10 d.to… Padrini… il S.r Bortolo Francal Giudice et la S.ra Madalena moglie del Sig. Vic.o Quelis». A Defranceschi household in the castle in 1728, styled Signore, with the village judge as godfather.

The method now scales. Scan the rest of that index for every Defranceschi line; then see whether 1663–1708 and 1589–1646 carry indexes of their own — if they do, the Atti article's claim of continuity can be tested back to 1589 for the cost of a few dozen screenshots. Marriages 1762–1858 has not been touched.

A second household, and a man followed across thirty-two years. The index's D page gives «Dom.co Gasparo di Giacomo Defranceschi 9 Ag.to 1732 — c. 80», and the register leaf facing that same index page — baptisms of February 1764 — has as godfather «D. Gasparo de Franceschi figlio di Giacomo». The child of 1732 standing at the font himself at thirty-one. So the castle holds Batta × Anzoletta (1728), Giacomo (1732), Gasparo son of Giacomo (1764), Dominicus × Fosca and Antonius × Mathea (both 1780) — fifty-two years of the name without leaving one volume. Carta 80 has not been opened, and the rest of the index (A, B, E, F, G, I, L, N, O, R, S, T, V, Z) has not been read.

Index letters C–G now read. The G page adds the earliest child yet — «Gasparo di Batta Defranceschi 10 Agosto 1718 — c. 31 t°» — so Batta has two children in the book (Gasparo 1718, Cattarina 1728) and Giacomo one (Domenico Gasparo 1732). Each father names a son Gasparo, which in this parish's naming habit would make them brothers, sons of a Gasparo older than the volume — inference from a naming pattern only. Still to read: index letters A, B, I, L, M, N, O, P, R, S, T, V, Z, and cartas 31 t° and 80.

B page: a third child and the father's name in full. «Battistina fig.a di Gio Batta Defranceschi 23 Xbre 1731 — c. 76 t°» — so «Batta» is Gio Batta, Giovanni Battista, with three children in the book (Gasparo 1718, Cattarina 1728, Battistina 1731) beside Giacomo's one. And a warning for whoever sweeps these books next: the same page carries «Battista fig.o di m.ro Piero Defranzieri 7 mag.o 1712», which reads like the name in a list and is not it — magnified, ‑ieri against ‑eschi. Dissolved. Letters read: B, C, D, E, F, G. Left: A, I, L, M, N, O, P, R, S, T, V, Z.

Sweep state, so it can be picked up cleanly. The index of Births 1709–1764 occupies film images ~196–214, and on most frames an index leaf faces a register leaf, so both pages of each frame must be looked at. Read: images 197, 198, 199, 200, 201, 202, and the right page of 207 — letters B, C, D, E, F, G and the 1720–23 stretch of M. Unread: 196, 203, 204, 205, 206, 207 left, 208–211, 212 (P), 213, 214 — letters A, I, L, most of M, N, O, P, R, S, T, V, Z, which is where Antonio, Domenico, Lucia, Maria, Nicolò, Pietro and Zuanne would be.

Three cartas are named by the index and still unopened: 31 t°, 76 t° and 80. Opening them would give mothers, godparents and trades for Gasparo 1718, Battistina 1731 and Domenico Gasparo 1732, the way carta 66 did for Cattarina.

A correction this item has to carry. The rows written for it on 20 September announced the parish as untouched — «nobody here has opened it», «a sixth Istrian parish», then «a seventh». All false. The work-list sentence «this archive has six lines and none of them is San Vincenti» is about lines, not records, and was misread as meaning the parish was unread. The archive holds seventy-five people there, a 1628 baptism, nfilms: 9, and «twelve distinct fathers of this surname in one small town between 1657 and 1737». What is genuinely new is everything after 1707.

And a conflict to settle before anything is merged. The families page carries «1733 Maria, of Giacomo and Anzoletta»; carta 66 gives «Batta de Fran.chi et di Anzoletta» in 1728. Two husbands for one forename. Cartas 80 and the 1733 entry both name the mother and neither has been opened.

20 Sept, later — L and the head of M read, and M paid. «Maria d' m.ro Nardo defranceschi, 13 Agosto 1709 — c. 13 t°» and «Marina di Piero Defranceschi, 21 Febraro 1710 — c. 6». Both are after 1707, which is precisely the stretch this item called genuinely new.

Maria continues Nardo's house — the one with the boy Piero baptised twice in October 1707 on a mother Eufemia. Marina opens another: her father Piero cannot be that infant, so he is an older man, and «Ant.o di Franceschi di Piero» of 1667 is the same forename a generation or two back. Open.

L is clean — ninety lines of Lucia and Lorenzo, 1711–1733, no De Franceschi. Recorded so it is not read twice.

A third, on image 205: «Maria di Batto Defranceschi, 13 Luglio 1716 — c. 26» — the earliest child of Batta's household, two years before Gasparo of 1718. Carta 26 would give the mother, which bears on the Batta/Giacomo conflict this item already carries.

Letters read: B, C, D, E, F, G, L, and M through 1717 — images 203, 204, 205. Left: A, I, the rest of M, N, O, P, R, S, T, V, Z — images 196 and 206–213. Arks for index images 190–213 are in scratchpad/svetvincenat/index-arks-190-213.lst.

Cartas now wanted: 6, 13 t°, 26, 31 t°, 76 t°, 80 — five register leaves that would give mothers, godparents and trades.

21 Sept — the index is read end to end, and it stops at P.

It occupies images 196–211 and runs B, C, D, E, F, G, L, M, N, O, P. Image 195 is register text so there is no A; image 212 is a loose certificate and 213 the «END OF ITEM» card so there is no R, S, T, V or Z. The earlier list of «letters left» assumed pages that do not exist.

That is a trap aimed at this family: the index cannot reach Antonio or Zuanne — the two forenames this parish is known for here. A nil from this index is not a nil from the register.

Seven entries found, five households. New fathers Nadal (Natale, not Nardo) and Piero. Full table on the search register.

And the Batta/Giacomo conflict is closed. Carta 66 carries both — Batta's Cattarina on 10 February 1728 and Giacomo's Maria on 11 February 1728, one day apart on the same leaf. Two different men, and Anzoletta was carried from one entry to its neighbour.

What is left on this item is no longer the index. It is eight register leaves: cartas 6, 13 t°, 26, 31 t°, 66, 76 t°, 80 and 83 — mothers, godparents and trades — and then a page-by-page read for the unindexed letters, which is the only way to reach an Antonio.

21 September, second pass: seven of the eight leaves are open, and they pay better than the index did. The reading is on the search register. Three mothers — Vittoria (Piero, 1710), Elena (Giacomo, 1728), Anzoletta (Giacomo, 1733); a sixth man of the name, «il sig.r Marcio de Franceschi», godfather twice in seven weeks in 1709–10 and structurally invisible to an index arranged by the child's forename; and «D. Nadal Defraneschi» godfather in 1732, four entries above a «Nadal Venir» on the same leaf.

It also cost a published claim. This morning's note called the 1733 Anzoletta a clerk's carry-over; carta 83 says her name in its own hand, first entry on the leaf, nothing to carry from.

And the index's page numbers are not safe. «Marina di Piero, 21 Febraro 1710 — c. 6» is carta 16. Carta 76 t° and carta 26 are now open too. 76 t° gives Battistina, born 18 and baptised 23 November 1731 to Gio Battista and Maria — a month earlier than the index said and a mother it did not give. Carta 26 runs 18 April to 22 June 1716 and carries no Defranceschi at all, so the index's «13 Luglio 1716 — c. 26» has nothing behind it. Three of the index's four page references have now failed.

The wives resolve into two men who each married twice — Gio Battista: Anzoletta then Maria; Giacomo di Pietro: Elena then Anzoletta. Anzoletta sits in both households twenty years apart, which is the whole of the confusion.

Still to open: carta 31 t° (Gasparo, 1718) — and it should be found by date, not by carta. Then the page-by-page read for the letters the index never had, which remains the only route to an Antonio.

The page-by-page read has begun, in the right volume.

Antonio's exemption is 1780, and 1780 is in the second births volume (1764–1791), not the one whose index was swept. Frames 81, 82, 83 are read — 21 April to late October 1780 — and they give «Matthea uxor Antonij Defranceschi» on carta 80 and «Dominico Defranceschi» as godfather on carta 82/83. No child of the name is baptised in any of them.

Do not resume at frame 84 — it is January 1763. Frames 81–83 are April–October 1780 and frame 84 jumps seventeen years backwards, on consecutive film frames (005497885_00295–00298). The film does not run in date order here, and both years were verified at magnification. Date every frame from its own heading; never infer one from a frame number. Arks for images 71–110 are in scratchpad/svetvincenat/vol1764-arks-71-110.txt, and the two-call-per-frame spread crop is recorded with its own warning. The 1781–1791 leaves have still to be found, and finding them now means probing for them rather than walking forward.

Still to do: the rest of 1764–1791, carta 31 t° of the first volume (found by date, not by carta), and the first volume's unindexed letters.

Seven spreads read, and the silence redirects the search.

Images 81–83 (1780) and 93–95 (Dec 1781 – Feb 1782) carry three Defranceschi and all three are godparents — Matthea, Dominico, Jusca. No baptism of the name in any of them. The reading is on the search register.

A godparent is an established adult, so the children were born earlier. The generation to hunt is 1750–1775 — the end of the first volume and the opening of the second — not the years around the exemption. That is where the next pass should go, and it is a different place from where this one was heading.

The 1750–1763 window is now bracketed. Image 160 is carta 143, October 1751 – February 1752; image 185 is carta 168, June–October 1761 and in Italian, not Latin; image 195 is the last register leaf before the index. At ~3.26 images a year that puts the whole window at images 149–195, about forty-seven frames. Both probes were nil. Arks are in scratchpad/svetvincenat/vol1709-arks-151-196.txt.

And the spread crop has now produced four false positives in one session — Dassanich, Blasan, Jurca uxor Antonij, and Defranzieri from the earlier sweep. It has never once confirmed a name correctly. It is a finding tool only.

And the index has already narrowed the read to eight forenames. It holds seven Defranceschi entries and all seven are 1734 or earlier, in a volume running to 1763. It covers B–P, so if children of the name exist in 1735–1763 they are called Antonio, Andrea, Anzolo, Rocco, Simon, Tomaso, Vito or Zuanne — and Antonio is the man the item exists for. The reasoning is on the search register. Image 165 is carta 148, 1754, and the rate there is about two images a year, not the 3.26 extrapolated from the 1734–1761 span — so use the anchors, not the average.

Blocked mid-read: the FamilySearch session dropped.

The waypoint call began returning 503 and the tab was redirected to ident.familysearch.org/en/identity/login/?state=…/search/film/005497885. This archive does not enter credentials, so the register work stops here until David signs in again.

Everything needed to resume in one move is written down. The next step is not the forty-seven-frame baptism sweep — it is the marriages, 1762–1858 (waypoint 9R28-YWT:391644801,392404501,392405601), which was being opened when the session went. A marriage register names the groom's father, and Antonio's marriage to Matthea would settle in one entry what the baptisms would take a week to circle. Marriages are also far thinner than baptisms, so it is a smaller book to read.

If the marriages give nothing, fall back to the baptism sweep with the anchors already pinned — images 160 (1751–52), 165 (1754), 185 (1761), 195 (last leaf) — and look for eight forenames, not for a surname.

Signed in again, and the marriage register was the right call.

The marriages 1762–1815 (the title card's dates; the catalogue says 1862–1858) are open for the first time. Carta 11, 9 February 1766: «Elena figlia di q. Mattio de Franceschi, nativa di questo Castello» marries Mattio Sandalezza of Albona. The reading is on the search register.

A ninth man of the name — and dead before 1766. One leaf of marriages reached the generation that seven spreads of baptisms could not. Continue here: front matter runs to image 5, register text starts at image 6, and image 15 is carta 11 (1765–66). Arks 1–45 are in scratchpad/svetvincenat/marriages-1762-arks-1-45.txt.

Eight marriage frames read, and the volume is twice the size the catalogue implies. It is at least 85 images, not the 45 a single waypoint call returns — ask for start=45 as well. 1762–1779 occupies images 6–44; 1780–1815 is 45–85, and there is no index at the back of the first part.

Date anchors now pinned: img 6 = Nov 1762, 7 = Dec 1762–Mar 1763, 15 = carta 11 (1765–Feb 1766), 16 = May–Jul 1766, 17 = Aug 1766–Jun 1767, 18 = Jul–Nov 1767, 19 = Nov 1767–Jan 1768, 44 = carta 39 (1778–79). The register switches Italian → Latin around image 16 and back to Italian by image 44.

Read and nil: 6, 7, 16, 17, 18, 19, 44. Still to read for a pre-1780 Antonio: images 8–14 and 20–43 — thirty-one frames. Arks 1–85 are in scratchpad/svetvincenat/marriages-1762-arks-1-45.txt.

22 September 2026 — the next volume is open, the gap in it is measured, and the first frame read paid.

Births 1709–1764 is film 005497885, 214 frames, waypoint 9R2H-RM8:391644801,392404501,392404901. Anchored by reading rather than by interpolation: image 126 is headed 1740, image 161 is headed 1752 — about four frames to the year.

The reason to read it is now measured rather than assumed. Of this archive's 114 Svetvinčenat people, the ark-backed ones run 1666–1737 and then stop: next an indexed burial in 1758, next an indexed birth in 1769. The three children already held from 1753, 1756 and 1759 were every one of them read off a page by eye. So the index stops and the family does not — and 1738–1752 has been touched by neither, which is images ≈118–162.

Image 161 read, both pages, 1752 entire. It gives Irena, daughter of Matteo Defranceschi and Francesca, born 11 February 1752 — in no index — and, twice over, a godmother entered as the wife of Jacobo Defranceschi. Two households of the name at the castle in a single year, which is the Atti article's continuity seen from inside the register instead of quoted.

Next: images 118–160 and 162, the same way. And the marriages 1762–1858 afterwards, because Antonio of the 1780 exemption married inside that book and a marriage entry names the fathers.

22 September, later — images 115, 116 and 117 read (October 1737 – May 1738), and the name is on all three. Twins to Jacobo de Franceschi and Anzola on 26 December 1737, in no index; a Joannes Baptista de Franceschi dead by that autumn, known from his daughter's entry; Jacobo godfather three times and Maria godmother three times inside eight months. Four frames of about forty-five are read and every one has carried the name, so the remaining forty are a read worth finishing rather than a gamble. Still to do: images 118–125 and 127–160, and 162.

22 September, third pass — images 118, 119 and 121 read (June 1738 – March 1739). A third household: Anna Euphemia, daughter of Pietro de Franceschi and Giacoma, born 25 July 1738, in no index. And a caution that governs the rest of this sweep: in ten months the register gives «uxor Jacobi de Franceschi» three different wives — Anzola, Agata, Jacoba — with Angela (= Anzola) alive on both sides of the other two. Either there are two Jacobos in the parish, or a godparent line cannot be trusted about husbands. Filed as method.

Read so far: 115, 116, 117, 118, 119, 121, 161. Still to read: 120, 122–160, 162. Image 121's left page has a loose slip lying on it from the filming and is not a clean nil.

22 September, fourth pass — images 120 and 122 read. Image 120 turns out to be a second exposure of 121's opening (book page 107), with the loose slips in different places, so that opening is complete and this morning's caveat is withdrawn; the slips are the priest's own Italian memoranda, records in their own right. And there are two Joannes Baptistas — one defuncti by October 1737, one standing godfather on 19 April 1739. Read: 115–122 and 161. Still to read: 123–160 and 162.

22 September, fifth pass — images 123 and 125 (June 1739 – October 1740). A fourth household: Batta (Battista) Defranceschi and Jacoba, daughter Euphemia born 1 October 1739, godfather D.nus Aloysius Spinotti — almost certainly the Joannes Baptista standing godfather that April. Angela is Jacobo's wife in 1737 and twice in 1739, which makes the Agata and Jacoba of 1738 the anomalies rather than her. And a warning filed as method: the year headings on 123 and 126 read 1759 and the entries beneath them are 1739. Read: 115–123, 125, 126, 161. Still: 124, 127–160, 162.

22 September — and an entry this archive already had turns out to belong to the tangle. A reading made on 21 September, from carta 66, image 77 of the same parish, under «Adi 5 Febbro 1728»:

«Catta[rina] fig.a leg.ma di S.r Batta de Fran.chi et di Anzoletta sua Consorte»

So in 1728 a Batta is married to an Anzoletta — and Anzoletta, Anzola and Angela are one name. From 1737 to 1759 the register gives Angela as the wife of Jacobo, and in 1739 it gives Jacoba as the wife of Batta.

Either the same woman was married first to a Batta and then to a Jacobo, or the clerks of this parish could not keep the two men's forenames apart — which is the second time this register has raised that exact doubt. Recorded as a connection, not a conclusion. The 1728 entry was read before any of this and is not being re-read to fit it.

28 September — image 127 read, and the route to the plates had to be rebuilt from scratch first.

⚠ Every ark list this item pinned is gone. scratchpad/svetvincenat/ held four files of pre-resolved arks — index images 190–213, births 71–110 and 151–196, marriages 1–45 — and the scratchpad is session-local. It was cleared between sessions and none of them survived. Frame numbers and film numbers were recoverable from this note; the arks were not. Record the film and the frame, never only the ark.

The route, written down properly this time, because it is three hops and none of them is guessable:

1. GET /service/records/storage/dascloud/das/v2/dgs%3A<FILM>_<NNNNN> returns {"apid": "TH-…"}. Film 005497885, frame 00127 → TH-1942-28339-4054-72.
2. Navigate — not fetch — to https://sg30p0.familysearch.org/service/records/storage/dascloud/das/v2/<apid>/dist.jpg. A cross-origin fetch from familysearch.org returns 401; the browser following the redirect with its own cookies does not.
3. That lands on a signed S3 URL, good for one hour (X-Amz-Expires=3600), which curl will serve at full resolution — 5991 × 4402 — where curl against FamilySearch itself is refused by their security service.

Image 127 is carta 102, headed 1740, and carries the name. Die 25 9bris 1740, baptism of Andreas, son of Thomas Sirolich and Maria: «Patrini fuerunt Jacobus Defranceschi; et Maria uxor Dominici Rencoda».

Found on the spread crop and confirmed at 3× magnification before being written down — the surname wraps across a line break, «Jacobus» ending one line and «Defranceschi» opening the next, which is exactly the shape that has produced this item's four false positives. This one holds. It puts Jacobo at the font in November 1740, extending a man already known here in 1737, 1738 and 1739.

Still to read in this volume: images 124 and 128–160, and 162.

Image 128 read, both pages — a clean nil, and the first frame of this sweep to cost its full price.

Left leaf is carta 112 running into 1741; right leaf is carta 113, headed 1741, December 1740 to 25 March 1741. Every family on it is local — Chersan, Doblanovich, Doblolavich, Stancovich, Sulich, Bavcarich, Marcusin, Bassich, Jedreicich — and no form of this surname appears in any role.

⚠ The cost per frame is now measured, and it governs whether this sweep is the right use of the register. Each frame needs: one call to turn the film-and-frame into an apid, one navigation per frame to mint a signed S3 URL, a download, a crop, and one or two image reads. The signed URL cannot be reused across frames — its X-Amz-Date is per request, so reconstructing a second URL from a first one's query string returns 403, which was tried and failed. Call it five round trips a frame.

Frames read so far in the 1738–1752 gap: 115–123, 125, 126, 127, 128, 161. Still unread: 124, 129–160, 162 — thirty-four frames, so something like a hundred and seventy round trips at the present rate.

The marriage register remains the better instrument and this item said so before the sweep resumed: a marriage names both fathers outright, and Antonio's marriage to Matthea would settle in one entry what these baptisms circle. Its film number was not recorded — only a dead owc waypoint — and the two adjacent films (005497886, 639 images; 005497887, 634) do not obviously carry it. Finding that film is one afternoon's work and worth more than thirty-four baptism frames.

Images 124 and 129 read, and both carry the name — one of them a forename the index could never have reached.

124 — carta 110, headed 1740. Die 9 Februarij 1740: «…baptizavit unum infantulum natum die 6 currentis à Petro Bilich, et Joanna ejus legitima uxore ambobus ex hac Parochia, cui positum fuit nomen Joannes. Patrini fuerunt Petrus Defranceschi, et Marina Stancovich.» Pietro at the font in February 1740 — the man whose daughter Anna Euphemia was baptised in July 1738.

129 — carta 114, headed 1741, and it gives two. Right leaf, 27 August 1741: at the baptism of Michael, son of Joannes [Su]lcich and Maria, the godmother is «Francisca uxor Matthæi Defranceschi». That settles a wife rather than raising one: Matteo's wife at Irena's baptism in 1752 is also Francesca, so it is one woman across eleven years and not the two this register keeps producing elsewhere.

Left leaf, at the baptism of a Jacobus: «Patrini vero …anco, et Joanne Defranceschi». A Joannes — Zuanne — of the name, at Svetvinčenat in 1741. This is precisely the case the index could not reach: it runs B to P, so an Antonio or a Zuanne is invisible to it, and only a page-by-page read finds them. It is not Fiume's Joannes Baptista and is not offered as him — a Joannes is not a Joannes Baptista, and this parish reuses forenames freely enough that the match alone proves nothing.

Read so far in the gap: 115–129, 161. Unread: 130–160, 162 — thirty-two frames.

Images 130 and 131 — one of them untangles the wives, and the other is a lookalike dissolved.

130 — carta 115, headed 1742. Die 16 Februarij 1742, at the baptism of Matthia son of Gregorio Suficl(?) and Victoria: «Patrini ad sacrum fontem fuerunt, Joannes Solombat Paresin, et Jacoba uxor Petri Defranceschi.»

⚠ That is the tangle coming apart. This item recorded on 22 September that in ten months the register gives «uxor Jacobi de Franceschi» three different wives — Anzola, Agata, Jacoba — and filed it as a caution that a godparent line cannot be trusted about husbands. Here Jacoba is not Jacobo's wife at all: she is PIETRO's. Which agrees with the baptism already read on image 121 — Anna Euphemia, daughter of Pietro de Franceschi and Giacoma, 25 July 1738, Giacoma being the same name. So one of the three «wives of Jacobo» was never his, and the doubt shrinks from three women to two. (inf — the same forename in one small parish is not proof of the same woman, but Pietro × Giacoma is independently attested by his own child's baptism.)

131 — carta 116, headed 1742: a nil, and a near miss. «Patrini fuere Gregorij Ronsil ex Dabrova, et D.no Donato Del Ferri» — read at 3× and it is not this name. That is the fifth lookalike this sweep has dissolved — after Dassanich, Blasan, Jurca uxor Antonij and Defranzieri — and the first four were all found the same way, on the spread crop, and all four died at magnification.

Read in the gap: 115–131, 161. Unread: 132–160, 162 — thirty frames. Strike rate so far: six of seventeen frames carry the name.

Images 132 and 133 — both nil, and the volume changes its hand.

132 is carta 117, headed 1742, October to December: Milovcon, Kerkovich, Doblanovich, Lanich, Solombat, Berchiavaz, Crosciovaz, Brotolich, Drobilich. 133 is carta 118, headed 1743, February to May: Bulcich, Argelich, Vehovich, Contossich, Milovan, Lepieri, Ferlan, Smeglian, Bergomo, Milcevich.* No form of the surname on either.**

A change worth noting for whoever reads on. At carta 118 the clerk switches from the priest-first formula — «R.dus D.nus N. baptizavit unum infantulum … à Petro Bilich et Joanna» — to a child-first one: «Gregorius filius Simeonis Bulcich et Catherinæ ejus legitimæ uxoris». The surname of interest now falls in a different place in the sentence, which matters when the page is being scanned rather than read, and is the kind of thing that turns a sweep into a set of false nils.

Read in the gap: 115–133, 161. Unread: 134–160, 162 — twenty-eight frames. Six of nineteen have carried the name.

28 September 2026 — frames 134 and 135 re-read, and the tally above was wrong.

First, the tally. This item has been counting «unread: 134–160, 162 — twenty-eight frames». They are not unread. Frames 134 to 162 were read in earlier passes and already carry about twenty published people between them — Casparus 1743, Valentino 1745, Antonia 1745, Maria 1745, Martina Antonia 1746, Victoria 1746, Antonia 1747, the Vittorio twins of 1748, Euphemia 1749, Victoria 1749, Victoria 1750 and more. What frames 115–133 were was gap-filling; what 134 onward is, is verification. The running count was measuring the wrong thing, and said «twenty-eight frames to go» about a stretch that was already mined.

Second, what verification found on the very first frame of it.

Image 134, carta 119 — the mother of Casparus is Pascasia, not Francisca, and the godfather is Francesco Saverio Profumo, chancellor of the parish, not «Dominicus Xaverii Bosponi». Two words in one entry, both wrong, both published for five days. It matters beyond one child: the 1743 line was load-bearing for the «two Jacobos» conclusion, and with it re-read, 1743 no longer contributes to that case at all — there is one wife that year, not two. The conclusion survives on the 1752–53 evidence and is flagged as resting there. Full reasoning on /corrections/.

How it was settled, since the transcription alone settled nothing. Three words of the same clerk, at one magnification, side by side: his «Franciscus» on the facing leaf (carta 119v, 8 Oct 1743), his «Pascasia» nine images on (image 143, 31 Mar 1746), and the word in dispute. The disputed word is the 1746 word stroke for stroke and shares two letters with «Franciscus». A hand is only legible against itself — reading a single word at 8× and thinking hard about it is what produced the error in the first place.

Third, two readings this session withdrew before publishing them. On image 135 the two godmother lines were read at 3–5× as «Bascovicia uxor Jacobi Defranceschi», twice — which would have collapsed Pascasia and Marcario into one woman and demolished the two-men case outright. At 7× with contrast they are the published «Pascasia» (24 November) and «Mar-/:cario» (1 December), and the withdrawal stands. The date «23 9bris» read off the same crop was also wrong; it is 24, as published. Three false readings in one session, all from the same cause — a faint page and a pattern freshly in mind from the frame before — and all three killed by magnification before they reached a reader.

Image 135's left leaf (carta 119v, Aug–Oct 1743) is genuinely new and carries no form of the surname — Giulich, Benche, Marsani, Sloncovich, Jaricho, Varnier, Solombat, Cardin — but it is the leaf that supplied both controls: Profumo the chancellor with his wife Veneria, and the clerk's «Franciscus». A nil frame is not a wasted frame.

The verification pass, and a way of choosing which frames deserve it.

Twenty-one published readings rest on frames 134 and later. Re-reading all of them at magnification is days of work, so the question was which ones to re-read first — and the archive turned out to have been keeping the answer in its own prose. Seventeen of the twenty-one say «magnified» or «read letter by letter» in their notes. Four do not.

The four were re-read first. The result:

• Image 134 — Casparus 1743. Two errors (mother Francisca → Pascasia, godfather Bosponi → Profumo), plus the chaplain Zolonta → Zolonda. This is the one whose note never claimed magnification.
• Image 161 — Irena di Matteo 1752. Correct. «Francisca» is written with the clerk's tall crossed F and narrow minims; set beside «Pascasia» at 7× the two share not one letterform. Matteo × Francisca and Jacobo × Pascasia are two different women, provably, in one image.
• Image 148 — Dominico di Vittorio, twin 1748. Covered, not unchecked. He and Lucia are named in a single act — «Lucia, et Dominicus filii gemini Victorij Defranceschi, et Dominica ejus legitima uxoris» — and Lucia's note says «Magnified». The twins cannot have been read at different magnifications. His page lacks the word only because it is the sibling's cross-reference.

And two of the seventeen that do claim magnification were re-read as controls:

• Image 143 — Martina Antonia 1746. Correct. · Image 147 — Antonia 1747. Correct.

So the claim was worth something: three of three that asserted magnification held, and the one that did not asserted nothing and was wrong twice. That is a small sample and it is not a guarantee — but it is a defensible order to work in, and it is cheaper than re-reading twenty-one frames blind.

Verification pass: 134, 135, 143, 147, 148, 161 examined. The four-item risk set is cleared. What remains is the seventeen that claim magnification, which are now the lower priority — and the marriage register, which has never been opened at all.

28 September 2026 — the marriage film is found, and it was never a catalogue problem.

This item has carried «the marriage register — film number never recorded, only a dead owc waypoint» as a blocker for days. It is resolved, and not by asking anyone: every one of these films opens with its own target board, photographed at the Croatian State Archives in 1997, and the board states the register, the years and the item number in plain capitals.

Read off the boards themselves:

Both boards give operator Kekić Anita, project CRO 00009, filmed 17 and 19 June 1997.

And film 886 does not stop at 1623. Frame 300 is a printed Austrian «Liber Copulatorum» ruled in columns — sponsi, sponsae, parents of both, witnesses — and dated 1887. So the marriage film runs from 1581 to at least 1887 across several items, and the 1762–1858 volume the catalogue lists sits somewhere between frames 50 and 300.

⚠ What this means for the two Jacobos, and it is not good news. *The catalogue gives Svetvinčenat marriages as 1581–1589 / 1622–23 and 1762–1858. If that is right — and item 1's board agrees with its first half — then the marriages of the households this archive has just spent a week reconstructing do not survive. Jacobo × Pascasia, Matteo × Francisca, Pietro × Jacoba all married in the 1730s or earlier. There is no Svetvinčenat marriage register for those years to find them in. (The gap is the catalogue's claim plus item 1's board; the frames between 50 and 300 have not been walked, so it is not yet proved from the film.)

What the marriage film can still do is the next generation. The children baptised 1743–1760 married around 1765–1785, which falls squarely inside 1762–1858: Casparus b.1743, Antonia b.1747, Victoria b.1750, Irena b.1752, the Vittorio twins b.1748. A marriage act names both fathers, which is exactly the evidence that would separate the two Giacomos retrospectively — through their children rather than through themselves.

Next on marriages: walk frames 50–300 of 005497886 for the item boards, find where 1762 begins, and read forward from there for the surname.

28 September 2026 — the marriage volume opens, and the surname is on its first page.

The 1762 volume begins at frame 43 of film 005497886 — a flyleaf, then page «1», then «Adi 5 Luglio 1762 S. Vincenti». There is no index, so this volume must be walked. Frame 50 is 1765, frame 80 is 1778, frame 150 begins a new volume at 1815 — about 2.3 frames to the year, four marriages to the frame.

Frame 43, third entry — «Adi 6 Novembre 1762 S. Vincenti». Domenico son of Sime Bulcich marries Gioanna daughter of the late Mattio Bulcich, and the witnesses are «alla presenza mia di And.a Venier q.m Nadal, di M.ro Gasparo de Franceschi, e di molte altre persone Testimonj».

«M.ro» is Maestro — a master craftsman, and this archive has not carried a trade for any Defranceschi of this parish before. (doc for the appearance; the trade is a style of address, not a named craft, and the register does not say which.)

⚠ Which Gasparo is not settled and is not guessed. Three men of the name could be alive in Svetvinčenat in 1762: Gasparo husband of Maria, attested 1744–1750 and then about 45; Gasparo b. 1737, then 25; and Casparus di Jacobo b. 1743, then 19. «M.ro» suits an established man better than a youth of nineteen, which points at the first — but «suits better» is not evidence and the entry names no father, no wife and no age.

Frame 43's first entry is a gift of a different kind. «Adi 5 Luglio 1762» — Giuseppe, son of the late Sig.r Apostolo Basilisco of Canfanaro and current Chancellor of this Castle, marries Antonia, daughter of the late Sig.r Francesco Saverio Profumi. That is a fourth independent reading of the surname this session corrected from «Bosponi» — in Italian, in a different hand, by a different clerk, nineteen years after the Latin baptism entries. And it dates his death: Francesco Saverio Profumo, who stood godfather to Casparus in 1743, was dead by July 1762, and his daughter married the man who took his office.

Frame 43 read. The seam that matters — the marriages of the children baptised 1743–1760 — runs from here to about frame 108.

Marriage sweep, frames 43–48 — 1762 to November 1764, about forty-five marriages, one appearance.

Read at a scale that shows a whole opening in one view — both pages at 820×1980, which is enough to spot the surname even where it is not enough to read the entry. Anything that looks like it goes to 5× before it is written down, which is how frame 43's Gasparo was confirmed and how five lookalikes died earlier in this item.

The parish's recurring surnames in these years, for whoever reads on: Percovich, Bulcich, Zadro, Venier, Mazzon, Justinaz, Angelich, Milcich, Stancovich, Jacovich, Chersan, Vidulich, Salambat, Xivolich, Rosar, Missan, Petrovich. Not one Defranceschi marries in the first two and a half years of the volume — which is itself worth knowing, because the parish has at least four households of the name baptising children in the 1740s and 1750s.

Every entry is signed by the same man, «D. Simon Zadro Arciprete», who interrogates the couples himself and names the witnesses in his own presence. The formula is fixed and the surname always falls in the same two places — the couple's own names near the head, the witnesses at the foot — which makes a nil frame quick and a hit unmistakable.

Swept: 43–48. The seam runs on to about frame 108 for 1790, and to frame 149 for the whole volume.

Marriage sweep, frames 49–53 — and frame 53 pays for the whole volume.

Frame 53 carries a marriage and a widow, and between them they date two deaths.

22 October 1765 — «Domenico Calluzzi q.m Pietro … con Pasqua Ved.a q.m Giacomo de Fran-/ceschi in 2di uoti». Pascasia, widow of Giacomo de Franceschi, remarrying. The marriage book confirms a couple the baptism book built, it dates the death of a Jacobo for the first time in this archive, and it gives the two-men question a test by arithmetic: any living Giacomo de Franceschi acting after 22 October 1765 cannot be Pascasia's husband.

9 February 1766 — «Con Elena figlia di q.m Mattio de Franceschi nativa di questo Castello», marrying Mattio Santalezza of the territory of Albona, resident in the district twelve years. This is the first marriage of a person of the name this archive has ever found, she is almost certainly Helena b. 5 August 1742 at twenty-three, and her father Matteo is dead — «q.m» — which brackets his death to 1752 × 1766.

A note on the hand. Partway through frame 52 the register changes scribe: Arciprete Simon Zadro's loose Italian gives way to D. Mattio Adamich, Economo e Vicario, who writes a smaller, rounder, far more legible hand and signs «L'Economo Adamich per l'Arciprete Zadro». The formula changes with him — «hà contrato il matrimonio per uerba de presenti» replaces «contrasse il matrimonio», and the banns clause grows longer. The surname still falls in the same two places, so the sweep method holds.

Swept: 43–53. Eleven frames, about eighty marriages, three appearances of the name.

Marriage sweep, frames 54–56 — and a patronymic settles what a style of address could not.

«q.m Ant.o ex hoc Castro» — son of the late Antonio, of this castle. Three Gasparos were in play this morning when the 1762 witness was filed as unassigned. Two of them are now excluded outright: the twin of 1737 and Casparus of 1743 are both sons of Jacobo, named as such in their own baptisms. What a Maestro-versus-Domino argument could never have decided, one patronymic decided in a line.

It also dates two things. Antonio, the witness's father, is dead by 13 August 1766. And if the witness is Gasparo × Maria — the only surviving candidate — then he is alive that day and dead by 30 March 1773, when «Maria relicta q.m Gasparis de Franceschi» remarries.

⚠ And a caution this item should keep. He is «M.ro» in 1762 and «D.no» in 1766 — and it means nothing, because the scribe changed in between. D. Mattio Adamich styles almost everyone «D.no». Reading social rank off an abbreviation is the same error as reading a surname off a spread crop, and this item has now made a version of that mistake twice.

Swept: 43–56. Fourteen frames, about a hundred marriages, five appearances of the name — one marriage, one widow remarrying, three witnessings.

Marriage sweep, frames 57–64 — and a correction this item owes itself.

⚠ First the correction. This item said the marriage register had never been opened. That was wrong. What was missing was the film number in this worklist — not the register. The archive already held Giacomo de Franceschi (m. 1772), read from frame 63; Giacomo de Franceschi, d. before 1771, from frame 62; and six 1773 witnessings by Antonio de Franceschi with his son Vittorio, from frame 64. Three of the eight frames in this batch were already published. The sweep confirmed all three at magnification and changed none of them, which is worth something — but it is verification, not discovery, and this item had been calling it discovery.

The one genuinely new act is the best of them. 10 June 1771 — «Dominicus de Franceschi … junctus est cum Victoria Trecento … Testibus Marco […] ac Victorio de Franceschi». A Defranceschi marrying with a Defranceschi standing witness, which this archive has not had before. Read as Dominico the twin of 1748 at twenty-three with his father Vittorio beside him — and flagged inf, because two other Dominicos and two other Vittorios are alive and the act names no father.

And a distinction the sweep forced. «Gaspare de Franceschi q.m Ant.o» in August 1766 against a living Antonio witnessing six times in 1773: two Antonios, and the archive now carries both as separate pages with the reason stated on each.

Swept: 43–64. Twenty-two frames, about 170 marriages, twelve appearances of the name — three marriages, one widow remarrying, eight witnessings.

28 September 2026 — the sweep stops at frame 64, and it stops on purpose.

FamilySearch served a «Pardon Our Interruption» page to the frame-lookup endpoint after about seventy rapid calls. It is a bot check. This archive does not work around bot checks, so the automated sweep ends here rather than being retried, throttled or disguised.

Nothing is broken. David's session is intact and the ordinary film viewer loads normally — the limit is on hammering das/v2, not on the account. The remedy is patience, not cleverness: resume later and more slowly.

The waypoint for the marriage film, captured from the viewer while checking, so the next session need not rediscover it:

https://www.familysearch.org/search/film/005497886
waypoint 9R28-RMN:391644801,392404501,392405401 · collection cc=2040054

Where to pick up. Swept 43–64 (Jul 1762 – Feb 1773). Next: frame 65. The seam runs to about frame 108 for 1790 and frame 149 for the whole volume — call it forty-four frames to the end of the useful stretch.

And a rate to plan with. Twenty-two frames took roughly seventy endpoint calls, three per frame: one apid lookup, one signed-URL fetch, one image download. A slower cadence is the price of the route, and the route is still far cheaper than the viewer.

28 September 2026 — the bot check cleared, the sweep resumed slower, and frames 65–71 gave a daughter nobody had.

Resumed at roughly a third of the earlier rate — one endpoint call every 1.3 seconds instead of as fast as the browser would go. No further interruption.

The find is Fusca de Franceschi, 17 November 1773 — «Marc'Antonius Jacomini filius Andreæ, Parenzensis, et Fusca filia q. Mathiæ de Franceschi, Parochiana hujus Ecclesiæ». A third daughter for Matteo and Francisca, beside Helena and Irena — and what ties her to them is the witness: «Mathias Santalezza», the surname Helena married into in 1766. A brother-in-law at the altar. Her baptism has not been found, and no birth year is printed for her.

Two corrections to this archive's own pages, both upward. Antonio de Franceschi was recorded as «named six times in ten months of 1773». He runs to at least April 1775 and is now on a dozen entries. And Maria's 30 March 1773 remarriage, known from an index, is now read at source — «Maria v.da q.m Gasparis de Franceschi».

⚠ A trade offered and refused. Three entries write «Campanaro» — bell-ringer — directly after Antonio's surname. It would explain his whole career as a witness, and the archive is not taking it, because «Zvanne Sain campanaro» appears six times in the same two years. Either the collegiate church had two, or the office moved, or the word is doing something else. Left open — this session wrote a method note against reading rank off an abbreviation, and then nearly did it again four hours later.

Swept: 43–71. Twenty-nine frames, about 230 marriages, twenty-six appearances of the name.

Marriage sweep, frames 72–78 — Oct 1775 to July 1777, and two questions close.

1 · The bell-ringer, settled by word order. This item left it open this morning because «Campanaro» trailed Antonio's name and might have attached elsewhere. Frame 72 puts it in front: «e del Campanaro Antonio De Franceschi», where «del Sig.r» would stand. Five entries now, and he was a bell-ringer — which explains his whole career as a witness: he was at the church anyway.

⚠ And so was Zvanne Sain, who carries the same title a dozen times across the same years. Two at once — and the register says why without meaning to: it marries people at S. Vincenzo, S. Rocco, S. Germano and S. Maria Maddalena. A parish with four churches has more than one bell. (inf.)

2 · Elena's death, bracketed from her widower's second wedding. 26 May 1777: «Ved.o Mattio Santaleza q.m Micho ha contrato matrimonio … con Maria figlia del q.m Ant.o Bolich». «Ved.o» is a widower, and he is the man Elena married in 1766. So she died between 1766 and 1777, at thirty-four or younger. No year is printed: the bracket is eleven years wide and Svetvinčenat's deaths before 1859 are on no film this archive can reach.

Swept: 43–78. Thirty-six frames, about 290 marriages, twenty-nine appearances of the name.

28 Sept 2026 — marriage seam read to frame 85. Frames 79–85: four findings, three nils named. Antonia di q.m Giacomo marries a Carnian (7 Jan 1779); Domenico remarries as «Ved.o p.mi voti» six days later, which kills Victoria Trecento 1771×1779; Fosca Placich is a second Fosca the archive nearly conflated; Antonio still witnessing 22 Nov 1779. All four are in the Svetvinčenat line, not only here. Next: frame 86 onward — the seam runs to about frame 108 for 1790 and 149 for the volume.

Read to frame 91, the end of 1782. Two findings — Gregorio Defranceschi q.m Ant.o (11 Nov 1780) and «Vittorio Franceschi», the «De» dropped by a new archpriest (29 Apr 1782) — plus four named nils. The bell-rope has passed to the Bratulich. Next: frame 92 onward; the volume runs to about frame 149.

Read to frame 93 (1783). ⚠ Frame 93 overturned this session's own reading of frame 86: «Capo Antonio de Franceschi», twice, alive, where the morning had published «q.m Ant.o», dead. Correction recorded and removed from every file. The same frame marries Vittorio to Eufemia Bergamo at Parenzo, names Vittorio's father in a document for the first time, and puts evidence under an empty stub the family tree had carried for years. Next: frame 94 onward.

⚠ 28 September 2026 — paused, and not by choice. The in-app browser's FamilySearch session signed out mid-sweep: frame 93, read twenty minutes earlier, now returns 403 Access to artifact denied, and so does frame 94. This is David's to restore — sessions do not enter credentials. The sweep resumes at frame 94; frames 79–93 are read, written up and pushed.

Resumed 29 September (David restored the FamilySearch session). Read to frame 96 (mid-1784). Two findings: Zuanne figlio del q. Giacomo de Franceschi, the first son of Jacobo's house the register names, and Antonia «nata de Franceschi» dead 1779×1784, from her Carnian widower's remarriage. The Carnian count off this film stands at five in ten years. Next: frame 97 onward.

29 September — frames 80 and 97 read; the range is now contiguous 79–97. Frame 80 closed a gap the original sweep had left and only surfaced when page ranges were written up for Record Atlas — carta 37, June–August 1778, nil. Frame 97, carta 54, nil; 1784 closes at 14 marriages. Next: frame 98 onward; the volume runs to about 149.

⚠ 30 September 2026 — paused again at frame 98, and again not by choice. The in-app browser's FamilySearch session is signed out: the frame route returns {"error":"Unauthorized","message":"Unauthorized (anonymous session)","status":401} for dgs:005497886_00098. This is David's to restore — sessions do not enter credentials. Frames 79–97 are contiguous, read, written up and pushed; the sweep resumes at frame 98 and the volume runs to about 149. This is the second sign-out mid-sweep on this film, so it is worth saying plainly: the useful unit of work here is a short run of frames written up and pushed before the session can lapse, which is what the last two sittings did.

30 September — David restored the session; frame 98 read, and it is the best single entry this film has given. Carta 55, 1785. «Antonio q. Leonardo de Franceschi da Gollogorizza Diocesi di Pedena» marries «Maria figlia q. Giacomo Corva da Muina Diocesi d'Udine», 14 March 1785. That is the Muina household's missing marriage — the chain had inferred «about 1785» off a grandson's birth year and the year is right — it names Maria's father for the first time, it pushes Leonardo's death back before 1785, and it styles a man this archive has always called a farmer of Muina as being of Gologorica, the origin parish. The identification is inf and labelled so; the test is Osvaldo's 1788 baptism at Muina. Next: frame 99 onward; the volume runs to about 149.

Read to frame 100 (carta 57, 1786). Frames 99–100: fourteen marriages, all named, nil for the name as party, parent or witness — and the book totals the year itself: «Fine dell'Anno 1785 Matrimonii N. 16». Frames 79–100 are now contiguous. Next: frame 101 onward; the volume runs to about 149.

Read to frame 101 (carta 58, into 1787). Six more marriages, all named, nil for the name. And the volume proved the reading: it closes 1786 with «Matrimonii al Numero 10», and this sweep read exactly ten 1786 marriages across frames 99–101 — one on 99, eight on 100, one on 101. The parish's own count and this archive's agree, so no entry of that year was skipped. Frames 79–101 contiguous; next frame 102, the volume runs to about 149.

Read to frame 102 (carta 59, 1787). Seven marriages, all named, nil for the name. ⚠ And the registrar changes mid-page: from 18 February 1787 the entries are signed «Augustinus de Ceneda Can. et Oeconomus registravit», not D. Paolo Cossaich — a third hand in this volume. The last change of hand, in 1782, changed the family's surname from «Defranceschi» to «Franceschi» and is why an indexed search loses them mid-volume; the frames from here should be read expecting a third form, not scanned for one. Frames 79–102 contiguous; next 103, the volume runs to about 149.

Read to frame 103 (carta 60, into 1788) — and it is a hit. 16 January 1788: «Maria figlia del S.r Antonio Defranceschi» marries Zuanne Contossich q.m Pietro, witnessed by «Domenico Defranceschi». Two of the name in one entry. ⚠ And the «De» is back. Frame 102 flagged a new hand from 18 February 1787; here is what it did — D. Giovanni Venica performs and D. Agostino de Caneva registers, and the surname is written «Defranceschi» in full again after seven years of «Franceschi». The spelling tracks the scribe, not the family, and has now been watched moving both ways — which is the single most useful thing this volume has taught about searching it. Maria's father is styled «Signor»; that he is the Antonio of the 1780 exemption fits parish, period and honorific and is recorded as inf, not fact. Both written into the Svetvinčenat readings. Frames 79–103 contiguous; next 104.

Read to frame 104 (carta 61, 1788–1790). Eleven marriages, all named, nil for the name. The officiant now signs «Can.co ed Arciprete», and «Sig. Can.co Agostin Decaneva» appears as a witness — a sixth instance confirming the «de Caneva» correction. And the bell-rope moves twice more: Gionello 1785–86, Civitico 1787–90, written into the readings. Frames 79–104 contiguous; next 105, the volume runs to about 149.

Read to frame 105 (carta 62, into March 1791). Eleven marriages, all named, nil for the name. ⚠ One bride is «della Parochia di Carbune Austriaca» — Krbune, in the Austrian county of Pazin, where Venetian Svetvinčenat's catchment had not been seen to reach before, and which sits in the same cluster of villages as Gologorica and Gračišće. No connection claimed — recorded because this archive tracks that border. Frames 79–105 contiguous; next 106, the volume runs to about 149.

⚠ 30 September — THE MARRIAGE REGISTER ENDS AT FRAME 106, and this changes what «the volume runs to about 149» means. The last act is 7 March 1791; then a Latin note records the book produced at the bishop's second general visitation of 28 May 1791, inspected and returned, and the rest of the leaf is struck through. Frames 107 and 108 are LOOSE SLIPS bound in afterwards — twenty-six marriages, June 1791 to November 1792, one line each, no witnesses or publications. All read.

And one is ours: «Addì 30 [X]bre 1791 — … con Maria relita q.m Contossich NATA DEFRANCESCHI». That is the woman who married Zuanne Contossich on frame 103 in January 1788, widowed and remarrying inside four years. (inf, strongly held.) Told in the readings.

What remains of the film is no longer this seam. Frames 79–108 are contiguous and the marriages are finished to November 1792. Frames 109 onward are unexamined and are not assumed to be marriages — the next sitting should establish what series they are before reading them as one.

Frame 109 is a SECOND EXPOSURE of frame 108's slips, with the «BLEED THROUGH» backing card moved — so each exposure makes a different part legible. Two entries unreadable on 108 are plain on 109 (Iure Mazzan × Eufemia Smosina, 30 Sept 1792; Matte Paicovich × Oliva Petercich, 26 Oct 1792), and the «Maria relita q.m Contossich nata Defranceschi» entry appears on both — the hit is confirmed on two independent photographs. ⚠ Caution recorded: where a backing card is in use the leaf is thin enough for the reverse to read through, so anything faint on these slips may be text from the other side. Frames 79–109 now accounted for.

Frame 110 — the slips run to May 1794, and give Jacobo a daughter. «Addì 19 Maggio 1794 — Domenico Martincich incontrò matrimonio con Pasqua figlia del q.m Giacomo Defranceschi». That is a third named child of that house, with Antonia (1779) and Zuanne (1784). ⚠ And the archive already holds a Pasqua who was the WIDOW of a q.m Giacomo in 1765 — a daughter named for her mother is the obvious reading and is recorded as inf, against the archive's own finding that there were at least two men called Jacobo here. Told in the readings. Fourteen marriages on the frame; thirteen nil and named. Frames 79–110 accounted for.

Frame 111 — slips to October 1795, fifteen marriages, nil. One bride is «Maria figlia del q. Giacomo…», the same shape as the Pasqua entry, so the surname was magnified before the frame was called nil: it reads Godichia. And the catchment reaches Pisino — the district town of Austrian Istria, and the district Gologorica and Gračišće sit in. Frames 79–111 accounted for.

Frame 112 — slips to April 1797, sixteen marriages, nil. Two things kept: the scribe hedges a surname in his own hand — «Gherso Doblanovich, ò forsi Bullesich», «or perhaps Bullesich» — a clerk conceding he does not know which of two houses a man belongs to, which is the same problem this archive meets from the other end; and two households exchanging spouses in a single afternoon, 24 April 1797. Lupoglav makes a fourth Austrian-Istrian parish in this stretch. Frames 79–112 accounted for.

Frame 113 — slips to November 1798, sixteen marriages, nil. One entry kept for the archive rather than for this family: «Il Sig.r Giuseppe Fackinetti incontrò matrimonio con la Sig.ra Giuseppina Lagadopoli da Spalatro, ed esso da Montagnana» — a Greek-surnamed bride of Split marrying a Facchinetti at Svetvinčenat in January 1798. Facchinetti is a name this sweep already knows from frames 102–103. Not ours, no connection claimed. Frames 79–113 accounted for.

Frame 114 — slips to May 1802, sixteen marriages, nil. ⚠ The slips are not strictly chronological: a header «Addì 24 Xbre 1800» is struck through and the entry beneath it is dated 23 February 1801 — the keeper writing up out of order and correcting himself. Worth knowing before anyone treats these slips as a series. Also: a physician's son, «Il Sig.r Marcello Manari figlio dell'Ill.mo Sig. D.r Girolamo medico fisico», marrying a lady of Lindar; and Mune — a fifth and sixth Austrian-side parish after Carbune, Pisino and Lupoglav. Frames 79–114 accounted for.

⚠ 30 September — FRAME 115 FORCES A CORRECTION TO THIS ITEM'S OWN CLAIMS.

Frame 115 is frame 114's slips photographed again. Fifteen of sixteen entries match. One does not: where 114 shows the Manari marriage, 115 shows «Addì 24 Xbre 1799 — Zuane Bercan q. Giure … con Elena relita q. Tomaso Gaugnelin della Parrochia di Valle», an entry 114 concealed. And on the right-hand slip the mechanism is visible in the photograph: a smaller sheet lies on top of a larger one, edge and shadow plain, carrying a note of witnesses over the entry beneath.

So the slips are a STACK, not a set of sheets — and the same thing explains frames 108 and 109, which this item read as a backing card being moved. It was the pile being shifted.

Every slip row is re-filed walked: false. «Walked» means every entry read, and that cannot be claimed of a layered pile photographed once. «Walked» now stands only for the bound register, frames 79–106, where the pages are sewn, wholly visible, and the parish's own year-totals for 1785 and 1786 were checked against the count.

The four hits are unaffected — each was read off a clearly visible entry, and the Defranceschi widow of 1791 appears on two exposures. What is weakened is the nils: the slip frames say nothing found in what could be seen, not nothing there. Whoever finishes this film should obtain every exposure of each slip opening and compare them.

Branko Fučić's Glagoljski natpisi (JAZU, Zagreb 1982) is the corpus of Croatian Glagolitic inscriptions — the reference work, not a byway. It contains this: «Natpis je vjerojatno registriran u Gologorici, gdje su Defranceschi…» — the inscription was probably recorded at Gologorica, where the Defranceschi…

Gologorica is one of this archive's own origin places — the Gologorica and Gračišće line, whose registers open in 1644. A Glagolitic inscription connected to the name, at that village, in the standard scholarly corpus, and there is not one line about it on this site.

Why it could matter more than a register entry. Glagolitic is the Croatian liturgical script, and an inscription is stone or a book-margin note — it is older than any parish register in this archive and it is in the other language. A family that appears in Italian on Venetian paper and in Glagolitic on an Istrian wall is the whole argument of the name page in one object.

Most of the sentence has since been recovered, by walking the full-text index phrase by phrase — each search re-centres the snippet, so the sentence can be read a few words at a time out of a book whose pages are closed for copyright:

«Riječ »Pomjane« (ime istarskog sela) ne može se čitati u posljednjem retku. Natpis je vjerojatno registriran u Gologorici, gdje su Defranceschi imali posjed i …»

— «The word »Pomjane« (the name of an Istrian village) cannot be read in the last line. The inscription was probably registered at Gologorica, where the Defranceschi had an estate and …»

Two things fall out of that, and the second is the bigger. Pomjane is a village name this archive has never held — it is the word Fučić could not read on the stone. And «imali posjed» is not a mention, it is a landholding: Fučić places the family as proprietors at Gologorica, and dates the inscription by that fact rather than the other way round.

What is still needed is the last clause and the entry's apparatus — the inscription's catalogue number, its date, its reading and where the stone or page now is. The walk stops at «posjed i»; the rest is in the printed book. NSK, Glagoljski natpisi, Fučić, JAZU 1982 — a reading-room hour, or a reproduction order from digitalna.nsk.hr. Walked 20 September 2026.

20 Sept — dated, and deflated. Gologorica's Glagolitic record: an inscription on a discarded holy-water vessel base of 1549 at the All Saints chapel; graffiti of the 15th–16th century on the north wall of the church of the Blessed Virgin Mary kod lokve; and the 1595 inscription of the priest Juraj Kaligarić in the north chapel of SS Peter and Paul.

But nothing found says the family is named in any of them. The fragment that opened this item reads as though it did; the likelier completion is «…where the Defranceschi had their house» — the family as the place, not the text. Do not claim a Glagolitic inscription naming the name until Fučić's own entry is read.

What it gives instead is worth more. Gologorica was writing Glagolitic from at least 1549, and this family arrived into it and in 1711 cut arms, a Latin inscription and a relief of Christ over a door the encyclopaedia calls «a modest palace of village landowners». That is their social position, in one sentence, on the house itself.

Reading the marriage film for a Defranceschi produced a Kovačina instead. Crikvenica marriages 1895, entry 28, 28 November: «Petar Kovačina, stolar» — a joiner — marries «Marija Car», at Crikvenica house 119. His father is another Petar Kovačina.

Uršula Kovačina married Anton Rudolf Defranceschi and is David's great-grandmother; she was born at Crikvenica about 1890–95 and died at Belgrade in 1975. Until now her family existed in this archive only as a surname on a tree. Here is a Kovačina household in a register, with a trade and a house number.

What to do, and it is already half-ordered: the Crikvenica birth register 1890–1896 — wanted on the errands for the modern households anyway — would show whether Uršula is a daughter of this house, and would give her parents outright. House 119 is the thread to pull.

inf — a Kovačina at Crikvenica in 1895 is very likely her kin; which kin is exactly what is not yet known.

Done, 20 September 2026 — and it closed a question this archive had deliberately left open.

The Crikvenica marriage of 28 November 1895 gives the groom's parents in full: «Petar Kovačina, k. r. financ. nadstražar, i Uršula rodj. Stojčić» — the very two people the 1904 death entry names. Petar Kovačina the joiner is the son of Petar Kovačina the financial guard, not a brother and not the same man under an earlier trade.

Three generations of house 119 now stand: the guard and Uršula née Stojčić → the joiner and Marija Car (m. 1895, by leave of the Royal Hungarian ministry) → Uršula, who married Anton Rudolf.

What is still inference: that Anton Rudolf's wife is the joiner's daughter. Next: Crikvenica births 1896–1902 for a Uršula Kovačina at house 119.

Found on 20 September in the Crikvenica marriages, 1899 entry 12: «Josip Kalanj, zidar, iz Ledenica», of Ledenice house 37, son of Josip Kalanj and Rosalija née Perković, marrying Marija Poštrenčić, a seamstress. Those are Anna Kalanj's parents, from her own Ledenice baptism of 1864 — so this is her brother, and Anton Rudolf's maternal uncle.

It is the first time the bride's side of the Klenovica marriage has been in a register at all, and it says what they did: masons.

Three things follow, and all are in Ledenice:

• the Ledenice births, about 1860–1875, should hold Anna and her brothers in one household — the whole sibling set this archive has never had;
• the Ledenice marriages after 1882 are the volume that would hold Anna's own marriage to Josip Defranceschi, and they are not on the film this archive has been reading — they must be asked for from the Matični ured;
• Luka Kalanj, son of Martin, mason, stood witness, and Ivan Kalanj married in the same register three weeks earlier to different parents. There is more than one Kalanj household, and telling them apart is what the Ledenice births would do.

This is the film paying for itself in a way that was not the plan: it does not hold the Defranceschi marriage, but it has handed over the bride's family.

Done, 20 September 2026 — and it answered more than it asked.

Anna Kalanj is the third of eight children of Josip Kalanj and Rosalija née Perković: Thomas 1858 · Martinus 1861 · ANNA 1864 · Lucia 1866 · Franciscus 1868 · Josephus 1871 · Maria 1874 · Joannes 1877. The FamilySearch index located the pages; each entry was then read on the register image.

Josephus, born 5 May 1871, is the «Josip Kalanj, zidar» who married at Crikvenica in 1899 — so the brother found from the marriage and the brother found from the baptisms are the same man, confirmed from both ends.

And the house is the find. The clerk writes it as «Ledenice in Klenovica N° 22» (1861), «Rljinovica N° 22» (1864), «Ledenice N° 22 in Rljenovice» (1866) and «Povile N° 22» (1871, 1877). Anton Rudolf was born at «Klenovica kbr. 22» in 1890 — the same house. See item 76.

Bribir, Vjenčani 1858–1900 (waypoint 9R28-L2Q:391644801,391644802,391654901) covers the whole window in which the marriage of Josip Defranceschi × Anna Kalanj must fall, and this archive's earlier shelf survey wrongly recorded Bribir's marriages as stopping in 1858.

It is now the only filmed Vinodol marriage register left unread. Crikvenica 1859–1908 is read end to end and clean; Sv. Jakov-Šiljevica 1859–1910 is read across 1882–1891 and clean; Ledenice stops at 1882 and Novi Vinodolski at 1859.

The method is settled: the left page of these Austrian tabular registers carries both spouses, their origin and their domicile, so one crop a page finds the surname.

Done, 20 September 2026. Bribir is read and clean.

1883–1890 complete — about four hundred marriages — and no Defranceschi in any of them. Year totals off the priest's own numbering: 1883: 57 · 1884: 45 · 1885: 64 · 1886: 41 · 1887: 45 · 1888: 25 · 1889: 22 · 1890: 24.

A catalogue trap worth knowing: the waypoint for Vjenčani 1858–1900 returns 597 images, because Bribir's births, marriages and deaths share one reel. The marriages sit around images 108–150; everything past about 250 is the death register.

With this, every filmed Vinodol marriage register covering 1883–1889 has been read and every one is negative — Crikvenica 1859–1908, Sv. Jakov-Šiljevica 1859–1910, and now Bribir. Ledenice stops at 1882 and Novi Vinodolski at 1859. There is nothing left on film. What remains is the letter to Rijeka — item 141.

This is now the likeliest home of the marriage, and it cannot be read from a screen.

Anton Rudolf's baptism of 20 January 1890 is in a Klenovica register catalogued as «Matične knjige, Ledenice / Klenovica», and its remarks column says the parents were at Klenovica only because they were staying with Ana's parents — «kod roditelja žene Ane» — at house 22, where the godfather Jakov Kalanj also lived. Klenovica is the bride's family's village. Marriages are normally celebrated in the bride's parish.

Neither book is on FamilySearch. Klenovica does not exist as a parish in the FamilySearch tree at all; Ledenice's marriage film stops at 1882. So the two registers most likely to hold this marriage — Klenovica's own marriage book and Ledenice's marriages after 1882 — have to be asked for.

Where to ask: the Državni arhiv u Rijeci holds the Vinodol parish books (Crikvenica's own volume is theirs: Povijesni arhiv Rijeka 275 (K4), inv. 110), and the Matični ured for Novi Vinodolski holds what the archive has not taken in. Ask for the marriage register of Klenovica and of Ledenice covering 1883–1890, and for any entry for Josip Defranceschi and Ana Kalanj in either.

This is David's letter to send, not a film to read.

20 Sept 2026 — a second thing to ask for in the same letter. The Crikvenica births for 1897–1902 are wanted too, for Uršula Kovačina, daughter of Petar Kovačina stolar and Marija Car of house 119 — Anton Rudolf's wife. The whole of filmed 1896 has now been read entry by entry and she is not in it, the FamilySearch index holds no Uršula Kovačina birth anywhere in Croatia, and the parish's birth register is not filmed past 1896. Her parents married in November 1895 and a sister was born about March 1897, so she falls squarely in the unfilmed years. Until that register is read, the line from the financial guard's son to Anton Rudolf's wife is inference and nothing more.

This is the join that would matter most, and it is one register away.

The index sweep of 20 September turned up a four-generation chain at Rijeka: Giuseppe Defranceschi × Francesca → Valentino b. 1772, who fathers twenty-one indexed children on Ursula between 1794 and 1805 → Franciscus b. 1797, who fathers four more to 1836.

The archive's own Fiume household is Valentino the Carnian, citizen of 1739, fifteen children 1737–1757. A Giuseppe whose son is christened Valentino about 1772, in the same town, is what a grandson looks like — but there is no Giuseppe among the fifteen.

Why that is not fatal. The fifteen were fourteen until 20 September, when Jacobus Vincentius was found in the index and then read on the page. The 6 September reading covered the dates it was given, not the whole book.

What settles it. The Fiume St Vitus baptisms for 1758–1775, on film 005497925 — the same film already anchored for this archive, where image 293 is 21 July 1757 and the volume runs on. A Giuseppe born to Valentino and Helena after 1757 would close the gap; his absence over eighteen years would open it. Either way it is a bounded read of one film at known frames.

If it joins, it joins a lot. The Rijeka family of 1794–1805 is twenty-one children this archive has never attached to anything, and they would hang off the Carnian household directly.

Narrowed the same day, and the window is now five years wide.

The far end is dated. The index carries «Valentinum Josephum Gregorium Defranceschi, f. Josephi, m. Francæ, 11/12 March 1776, Vitipolis» — so Giuseppe × Franca were baptising at Fiume in 1776, and their son is the Valentino who fathers the twenty-one.

Which dates Giuseppe himself. A man with a son born in 1776 was born by about 1750 — inside Valentino the Carnian's childbearing window, not after it. So the question is not «is there a sixteenth child after 1757» but «is there one in the gaps».

And the fifteen have gaps. They fall in 1737, 1739, 1741, 1742×3, 1745, 1751, 1752, 1753, 1754, 1755 and 1757×3. The holes are 1743–44, 1746–50 and 1756 — and 1746–1750 is a five-year silence in a household that otherwise produces a child every year or two. That is where a Giuseppe would be.

The read, precisely. Film 005497925. Ignatius was found at image 126 = 29 July 1745 and Vitus at image 228 = 18 June 1751, so 1746–1750 lies between them — roughly 95 frames, both pages each. ⚠ The frame arithmetic across that span is not safe: images 126 and 228 sit under different waypoints (9R2X-ZNT…714201 and 9R2X-C62…714501), and image 203 of the first shows a year in the 1780s, so the two are not one continuous sequence. Anchor by reading, not by interpolating across the join.

The index cannot do it. A sweep of 2,251 rows finds no Giuseppe with a Valentin- father anywhere, and four of the fifteen children were already known to be in no index at all. Absence in the index is not absence in the book — this item exists because that has already been proved twice on this film.

Scoped properly, and the earlier warning on this item was wrong.

This item said the two ends of the read sat under different waypoints and that image 203 of the first showed a year in the 1780s, so the frames could not be interpolated across. That was a misreading of image 203. Probing the volume directly shows it is continuous:

So volume one runs at about 8.3 frames to the year, and its baptisms stop before image 171, where a marriage register begins on the same reel. The window is images 127–170 — forty-four frames, not ninety-five, and there is no unsafe join inside it.

The index will not do it. There are no indexed Defranceschi baptisms at Fiume at all between 1746 and 1750 — but that discriminates nothing, because three of the children already known from this stretch (Ignatius 1745, Vitus 1751, Rosalia 1755) are in no index either. Silence in the index is this film's normal state for this family.

Ready to run: forty-four frames, both pages, looking for «Valentini» in the father's slot. The arks are in scratchpad/fiume/arks.lst, lines 127–170.

20 Sept, later — one page beat the forty-four. Giuseppe × Francesca have four children indexed at Fiume (1772, 1774, 1776, 1778). The eldest son's entry, 23 November 1772, reads «Valentinus Franciscus Xaverius, filius legitimus DD. Josephi, et Franciscæ jugalium Defranceschi, heri natus, baptizavit C.us Lumaga, Levantibus D.no Francisco Tomassich et D.na Helena Defranceschi».

Valentino the Carnian's wife was Helena, and in 1772 she would be in her mid-fifties. A Helena Defranceschi at the font for Giuseppe's first son is a grandmother. The celebrant, Canonicus Lumaga, is the same name as the canon who baptised Valentino's own son Aloysius Rochus in 1754.

Still inf — another Helena Defranceschi at Fiume is possible. The 1746–1750 read (images 127–170) would settle it outright and remains the way to close this item.

20 Sept, later still — the scoping above is wrong and the volumes are now mapped properly. Each waypoint on this reel is a bound volume holding births and marriages together, with different date ranges: 9R2X-ZNT = births 1734–1749 + marriages 1734–1753; 9R2X-C62 = births 1749–1769 + marriages 1753–1776. So «forty-four frames, images 127–170» was wrong twice over — volume one's births stop at 1749, and images 171+ are marriages of 1738. Image 203 is 1753, not the 1780s; that misreading has now been made twice and is retired.

Baptisms read so far: volume 1 images 127–133, both pages, August 1745 to May 1746. No Defranceschi as father. Incidentally Helena Defranceschi stands godmother four times in those seven frames, and a Catharina Defranceschi once — who cannot be Josepha Catharina b. 1737, aged eight. A Defranceschi woman not yet in this archive.

Marriages read: volume 2 images 286, 291, 292, 293 — March 1769 to December 1771. No Giuseppe. But 293R gives Joannes Baptista Defranceschi, widower, marrying Catharina de Benzoni on 9 December 1771.

Next, and cheapest first: volume 2 images 294–304 (marriages 1772–1776) for Giuseppe's own wedding, which would name his father outright. Then volume 1 images 134–170 and volume 2's early births for the 1746–1749 baptism gap.

Marriage sweep done, and it is a nil that redirects the item. Both pages of images 286, 289–295 of volume 2 — 1769 through August 1772 — hold no Giuseppe Defranceschi marriage. Since he baptises his first child at Fiume on 23 November 1772, he married outside the town, almost certainly in Francesca's parish. The entry that names his father exists; it is somewhere else, and the way to it is to identify Francesca. Images 287–288 (about June–November 1769) are the only gap left in the Fiume run.

And a caution on the Benzoni: they witness unrelated gentry marriages on image 289 too, so their appearance around Valentino's children in 1754–55 is not a family marker and should not be leaned on.

Baptism sweep tally. Both pages of images 127–134, 136, 137 now read — 22 August 1745 to 2 October 1746. No Defranceschi as a father in any of it. Unread: 138–170 (Oct 1746 to 1749) and the second gap at images ~104–125 (1743–44).

And the four children of Giuseppe × Francesca now have arks, so whoever picks this up can go straight to them: 1772 3QS7-899F-WCFK (read — Helena Defranceschi godmother); 1774 3QS7-899F-WHG6; 1776 3QS7-899F-WC34; 1778 3QSQ-G99F-WHGT (located on the page — «Die 8 Feb. 1778. Hellenam Ludovicam filiam legitimam D.ni Josephi et…» — godparent line not yet magnified). Their godparents are the cheapest remaining lead: a Defranceschi sibling standing for one of them would name the household without needing the 1746–49 grind.

The godparent read paid, and it is the best evidence this item has. Giuseppe's second child, 12 July 1774 (3QS7-899F-WHG6): «Hellenam Margaritham Mariam filiam legitimam D.D. Josephi et Franciscæ jugalium Defranceschi… levantibus D.no Ignatio Defranceschi et D.na Theresia Tomasich».

Ignatius Francus Mattheus is one of Valentino's fifteen, baptised 29 July 1745 on image 126 of this same film — one of the four in no index at all. In 1774 he is twenty-eight. So Valentino's wife stands for Giuseppe's first child (1772) and Valentino's son for his second (1774). The forename is rare and this archive knows no other Ignatius Defranceschi at Fiume.

Still short of proof — no entry says «Josephus filius Valentini». Remaining: the 1776 baptism (3QS7-899F-WC34) and the 1778 godparent line (3QSQ-G99F-WHGT), then the 1746–1749 gap or a death entry giving Giuseppe's age and father.

All four godparent sets now read. 1772 Helena Defranceschi · 1774 Ignatio Defranceschi · 1776 Carolo Müller and Josepha Callinesi — no Defranceschi · 1778 Ludovico Henrich and Hellena Carina — no Defranceschi (resolved on a fifth crop; Carina also witnessed the Defranceschi–Benzoni marriage of 1771). So: two of four, settled, and settled against the reading this archive wanted. Fiume godparents were as much patronage as kin.

20 Sept, the sweep resumed — and the recurring Helena now has a name. Both pages of images 138–140 read: 10 November 1746 to 17 February 1747, no Defranceschi father. The running total of volume 1 baptisms read is now 127–134, 136–140 — 22 August 1745 to 17 February 1747.

On 140R, 26 January 1747, «patrini fuerunt D.nus Cosar, et Helena Defranceschi» — her fifth appearance in fourteen frames. Valentino the Carnian's wife was Helena, and she is the only Helena Defranceschi known to be living at Fiume then, so the recurring godmother is almost certainly her. She is written plain in 1747 and D.na in 1772, matching the household's own rise from Valentinus to D.nus Valentinus in the 1750s.

This cuts both ways and the cut against is the important one. Five appearances in fourteen frames make her a habitual godmother, so her presence at Giuseppe's first son in 1772 is weak evidence of kinship — weaker than this item once claimed. Ignatio in 1774 is the load-bearing one, being a single appearance of a rare forename. And a rival Helena exists: b. 1725, d. 1809, of Gologorica — right age, wrong town. The entries give no patronymic, so the identification is by residence only. inf.

Death route closed for now. The index carries no burial for our Giuseppe — 36 rows of the forename, one death-typed, and that man was born 1784. A death entry would give his age and probably his father, so it remains the best single document; it just has to be browsed, in Fiume Deaths 1766–1815 (9R26-YWL).

Unread on the gap: volume 1 images 141–170 (Feb 1747 to 1749) and volume 2's early births for 1749; plus the second gap at images ~104–125 (1743–44).

20 Sept — the gap gave up a child, and it is not the one this item wanted. Images 138–146 of volume 1 now read, both pages: 10 November 1746 to 16 October 1747, continuous. On image 146L, under «Die 13 8bris 1747»:

«Franciscam Matthæam filiam legitimam D.ni Valentini, et Helenæ jugalium Defranceschi, natam 10 hujus, baptizavit D.nus Canonicus Hieronymus Moroldo.»

A sixteenth child — a daughter, born 10 October 1747, in no index anywhere. She is not Matheus Franciscus, who is 1739 and has his own ark.

This retires the premise of this item and sharpens it at the same time. The «five-year silence of 1746–1750» was the reason to think a hidden Giuseppe was plausible. There was no silence — the household was baptising in 1747, and the archive simply had not read the page. So the argument from absence is gone. But the same fact says the gap is productive: the first nine frames swept produced a Defranceschi baptism, and images 147–170 and ~104–125 are still unread. If a Giuseppe exists, that is where he is, and he will be found by reading rather than by inference.

A correction has been published for the «fifteen children» heading, for the five-year-silence claim, and for dating the household's D.nus / Ill.mus styling to «the 1750s» — the 1747 entry already carries both.

Helena's tally is now eight: four in images 127–133, then 26 January, 26 February, 26 March and 30 July 1747. That is a parish godmother, and it is the reason her 1772 appearance at Giuseppe's son carries little weight on its own.

20 Sept — answered, and not in the direction this item was pointing. Image 150R of volume 1:

«Die 29 May 1748. Josephum Aloysium filium legitimum D.ni Joannis Baptæ et D.næ Magdalenæ jugalium de Franceschi, natum 24 hujus, baptizavit D.nus Canonicus Eustachius Sauf. Patrini fuere D.nus Josephus Voxila et D.na Eleonora Franul.»

Giuseppe was born on 24 May 1748, and his father was Joannes Baptista, not Valentino.

The identification rests on five things that agree: the forename, the surname, the town, an age of 24 at his first child's baptism in November 1772 — and the fact that this archive had already found «Joannes Baptista Defranceschi, widower» marrying Catharina de Benzoni on 9 December 1771, which is exactly the status of a man whose wife Magdalena had died. The MyHeritage tree's unsourced b. 1748, d. 1778 for Giuseppe now has a document behind the birth year, and the death year sits in the same year as his last child's baptism.

Still inf: no entry says «Josephus filius Joannis Baptistæ». But this is far stronger than the godparent argument, and it is the first document that gives Giuseppe a father at all.

So the answer to the question as asked is: no. Giuseppe is not a sixteenth child of Valentino. He belongs to a second Defranceschi household at Fiume, and Helena in 1772 and Ignatio in 1774 were the other household standing for its kin — which was always the better reading of a habitual godmother.

Marking this done and opening the question it leaves. How Joannes Baptista and Valentino are related is now the join that matters, and it is answerable: their marriage entries are on this same reel, in volume 1's marriages of 1734–1753, and a marriage entry names a father and usually a birthplace. That is item 143.

Two Defranceschi households baptised at Fiume in the 1740s, and nothing yet connects them.

Valentino × Helena — «venuto dalla Carnia», citizen of Fiume in 1739, sixteen children 1737–1757.

Joannes Baptista × Magdalena — styled D.nus, baptising Josephus Aloysius on 29 May 1748 (film 005497925, image 150). A widower by 9 December 1771, when he marries Catharina de Benzoni (volume 2, image 293), with indexed children by her in 1773 and 1776. Josephus Aloysius is almost certainly the Giuseppe who heads the Rijeka chain, which makes this household the trunk of twenty-six indexed descendants, not a footnote.

He is not Valentino's son. The only Joannes Baptista among Valentino's sixteen was baptised in 1751 and cannot father a child in 1748. Brother, cousin, or no relation at all — the register has not said.

Why it is worth a name: if they are brothers, the Rijeka chain hangs off Carnia and joins the Mione line this archive already holds. If they are not, Fiume held two unrelated families of the name and a great deal of this archive's Kvarner material has been read as one thing when it is two.

The read, precisely. Film 005497925, volume 1, which carries marriages 1734–1753 from about image 171 — image 171 is already known to show marriages of 1738. Joannes Baptista × Magdalena must marry before May 1748, so the window is 1734–1748, and Valentino × Helena marry before November 1737, so they are in the same run or just before it. A Fiume marriage entry names the father of each party and usually where he is from. Two entries would settle it.

Cheaper still if it works: volume 2 carries marriages 1753–1776, already partly read for 1769–1772. Nothing there bears on this, so the read is volume 1's marriage section and nothing else.

20 Sept — one of the two marriages is found. Walking the marriage run backwards from image 189 (October 1746) through 170 and 169 to 168R, second entry:

«Die 28 Februarij 1737. D.num Valentinum de Franceschi ex Carnia, et Elenam Surcho junxit in matrimonium D.nus Petrus Eustachius Sauss Canonicus, præsentibus testibus Nob. D.no Petro Petrini et Nob. D.no Antonio Moretti.»

«Ex Carnia» is now primary, two years before the citizenship of 1739, and Valentino's wife has a maiden name for the first time: Elena Surcho. (Initial S or Z; given as read.)

But the entry names no father — Fiume marriage entries of this decade give the celebrant and the witnesses and stop there. So this join will not be made by filiation. It will be made, if at all, by finding Joannes Baptista × Magdalena in the same run and seeing whether the clerk writes «ex Carnia» over him too — and, better, by whether their witnesses are the same men.

Read so far: images 168, 169, 170, 189. Unread: 171–188 (1738–1746) and 190–205 (1747–1753). Joannes Baptista must marry before May 1748, so 171–190 is the window, and it is twenty frames.

20 Sept — the Fiume marriage run is now read end to end, and the answer is a nil that moves the item. Images 168–191, both pages, November 1736 to February 1748. In twelve years there is one Defranceschi marriage: Valentino's own. Joannes Baptista × Magdalena are not in it.

Their son was born 24 May 1748, so if they had married at Fiume they would be inside that window. They married somewhere else — exactly as their son Giuseppe would, and exactly as an incomer does.

So this item leaves Fiume. Kobler's claim is that a Giovanni Battista De Franceschi came to Fiume from the Contea di Pisino intorno al 1740 and was «senz'altro» of the Gologorica family — this archive's own line. His pedigree adds that the man studied at Fiume and married a lady of the D'Agnese family.

The test is now a single fact: Magdalena's maiden name. D'Agnese confirms Kobler and hangs the whole Rijeka chain off this archive's own line. Anything else kills it. And the place to find it is the County of Pazin's registers, which run from 1582 and this archive has never opened, not Fiume.

Incidental, and worth having: Valentino stands witness at a Fiume wedding on 28 February 1740 (image 175L), three years to the day after his own — an occurrence in no index. And a near-miss is retired: «Perill.ri D.no Josepho Defranchi», witness on image 187R in December 1745, is de Franchi, a different surname.

20 Sept, scoped — and the shelf is shorter than the question. Of the collection's 43 Istrian parishes, only two in the County of Pazin have marriage books covering 1740–1748: Pazin (1656–1852) and Tinjan (1703–1896). Gologorica, Gračišće, Pićan and Žminj are not in the collection at all — and Gologorica is the house with 1711 over the door.

So a nil at Pazin and Tinjan does not disprove Kobler. It would mean the entry is in a book that cannot be opened from here, which is an errand, not a read. Do not let a negative close this item.

Tinjan first. Darovec, quoted on this site already, puts the Friulian-Carnian stratum in «Pazin, Pićan, Gradišće, Žminj and Tinjan» — and Tinjan is the only one of the five that is filmed.

Pazin is enumerated and anchored, ready to run: waypoint 9R26-16F:391644801,392372801,392373301, 396 frames, arks saved to scratchpad/pazin/arks.lst. Index 134 = January 1741. Index 142 = 1746. Index 160 = October 1747 – January 1748. Read frames 128–165. The volume binds three date-runs together, so anchor by reading and never by interpolating across them.

20 Sept — Tinjan read, and it is a nil. Marriages 1703–1896, frames 9–13, both pages: 1733–1735 and 1741–1749. No form of the surname anywhere — not groom, bride, father or witness. The grooms are Červar, Milinarić, Antolović, Šubić, Smokovac, Dobrila: a Croatian village marrying its own, with Italian names appearing only as the notaries who witness everything.

That is worth having: it says an incoming Carnian family is not quietly marrying into Tinjan in these years, which the Darovec quotation might have led this archive to assume.

Gap: 1736–1740 was not located. Frame 9 ends 1735, frame 10 opens «Conjuncti sub Par.o Joanne… 1741». The volume already runs out of order, so those years may be elsewhere in it. A marriage up to 1747 is covered; 1736–1740 is not.

Still open, in order: Tinjan 1736–1740 if it exists in the volume; then Pazin marriages, frames 128–165 (arks in scratchpad/pazin/arks.lst). And neither can close this item, because Gologorica, Gračišće, Pićan and Žminj are not filmed here — reaching those is an errand, not a read.

⚠ 20 Sept, later — the Pazin frame window published above is WRONG. Do not use it. The years read as 1741 / 1746 / 1747 at frames 134 / 142 / 160 are 1711 / 1716 / 1717: this clerk's 4 and 1 are not distinguishable at page scale. Frame 143 writes 1717 twice cleanly, which cannot sit eight frames after 1746.

Re-anchored, all read: 143 = 1717, 178 = 1733, 198 = 1742, 200 = blank leaf, 205 = a microfilm resolution target, i.e. a volume boundary. ≈ 2.2 frames to the year.

The window for 1740–1748 is frames ≈194–204, and it runs into the boundary at 205 — so check whether the last years continue in the section after the target. Eleven frames and a join. Arks 175–234 are in scratchpad/pazin/arks175.lst; 0–179 in arks.lst.

A correction is published for this. When anchoring by date heading in this hand, cross-check every year against a neighbouring frame — and check that the run is possible before writing it down.

20 Sept — Pazin read end to end, and it is a nil. The online shelf for this question is exhausted.

Frames 194–199 (Dec 1740 – Nov 1742) and 208–216 (1743 – Nov 1748), both pages each. No form of the surname in any role. The couples are Sgrablich, Mattesich, Cerovaz, Brattulich, Sironich, Mogorovich — an Istrian town marrying its own — and the Italian and German names belong to the Praepositus and the chaplain who officiate.