Corrections

The Crete lane on the lines diagram carried a node reading «Wrecked · 1669 — Candia falls; Morosini's galleys take them off and the fleet is wrecked in a storm», and the lane's confidence was documented.

⚠⚠ FIVE SOURCES ON THAT EXACT EVENT WERE READ ON 12 OCTOBER AND NOT ONE MENTIONS A WRECK. The closest thing to an eyewitness account — the relation the Grand Master of Malta sent to the Viceroy of Sicily, published by Floristán in 2025 — describes the opposite: a negotiated embarkation stretched over twenty-two days, the Grand Vizier granting extension after extension while his army waited «con una quietud y una paciencia extraordinaria», and finally offering to forward anything left behind to Suda. About 12,000 people left Candia. Jackson (1887), Ortalli & Schmitt, Pérez Negrete's chapter on the war and the Ca' Foscari thesis on the defence of Candia are all silent on a wreck too.

⚠ And this archive's own spine never said it. On /seget/ the Candia stop reads «the twenty-one-year siege ends, and the Venetians take their people out» — which is exactly what the sources describe. The drawing and the spine were telling two different stories and only the drawing called its version documented.

What changed: the node is renamed «Taken off» and its note now carries the evacuation as the sources give it, with the wreck recorded as an open question rather than a fact. It is not deleted — an absence in five books is not a disproof, and it may be the Seget family's own tradition.

⭐ A possible origin, offered as suspicion and not finding: the Candia war does contain a Venetian fleet wrecked in a storm — Psara, the night of 17–18 March 1648, where seventeen galleys went down and the Captain General Grimani drowned. That is twenty-one years before the evacuation. A tradition could have slid from the one to the other.

⭐⭐ And the same reading gave the page what it had always lacked: why Istria at all. Morosini «ya había decidido su asentamiento en lugares deshabitados del golfo Adriático» — he had already decided the refugees' settlement in the uninhabited places of the Adriatic, and some who wanted to accept the Vizier's offer to stay could not. He did it while surrendering the city without the Senate's authorisation.

· Corrected in place: /lines/ · /seget/ · /searched/

⚠⚠ THE SCOUT OF THE DEATH REGISTER WAS CALLED «END TO END» AND IT WAS NOT. It read the top of every page and cut the bottom off.

The crop was y 0.09–0.46 of the frame. Measured against the book, the page itself runs from about 0.12 to 0.92. So roughly the top 44% of every page was read and the rest was never seen. Eleven people of the name were found, and they were real — but the claim that the book had been covered end to end was false.

This is the crop-drift trap this archive has now hit five times, and the previous four were all on the Pazin reels. It is the same failure: a crop that looks right on the frame you chose it with, and silently clips every frame after.

How it was caught. David asked whether «Medolino» was Medulin, and noted that Defranceschi seem to cluster there. Checking that sent this session back to a lead it had never taken up — XQM3-21M6, «Mattio Francesch, d. 20 Dec 1839, Ližnjan» — and asked why the scout had not reported that death. Because the entry sits at 0.52 of the page height, below where the crop stopped. A question about a place name exposed a systematic hole in a sweep.

The fix is one number. At the same magnification, y 0.09–0.95 renders the whole page into a single 2048×2304 canvas that reads exactly as well as the old one. The deaths book is being re-scanned with it, and the marriage and birth helpers carried the same wrong y-range and are corrected before use.

And the first thing the corrected crop found was the entry that exposed it:

Ližnjan «Liber Defunctorum», page 93, 1839:

> «Addì 20 Xbre 1839» · house 17 · «Mattio Franceschi d.to Marulin, agr.» · Cattolico, Maschio, «ann. 40»

— the death of the man who married in 1815, with «d.to Marulin» written again, twenty-four years after the marriage did. That settles the byname: it is attached to him twice, in two books, in two hands.

· Corrected in place: /searched/ · /liznjan/ · /worklist/

The roster carried «Josephus Defranceschi, b. 1887» from the FamilySearch index and «Josephus Defranceschi, b. 1886» from the MyHeritage tree. One child.

Page 99, entry 7 of 1888 — and the scribe wrote the name in the left margin as well as in the column:

> «Josephus Defranceschi» · «Mort. — Martii, Sep. — ejusdem, 1888» · house 82 · Anagr. p. 305, III. 9 · «Josephus Defranceschi, Gregorii et Maria Vojnić» · «an. 1, mens. 11» · «nat. 15/4 86»

Born 15 April 1886, dead in March 1888 at a year and eleven months — the age and the birth agree with each other, and the day/month/year form of «15/4 86» is the one this book uses throughout («nat. 2/5 851» against an age of «ann. 23, m. 8, d. 26»).

⚠ THE DEATH DAY IS BLANK IN THE REGISTER ITSELF. The scribe wrote «Mort. — Martii» and «Sep. — ejusdem» and never filled the days in. That is the book's gap, not a failure of reading, and no day is invented here.

⚠⚠ TWO ROWS, ONE CHILD — and this time the TREE was right and the INDEX was wrong. This archive held josephus-defranceschi-1887 from the index («b. Mar 1887») and josephus-defranceschi-1886 from the family tree («1886»). The register says 15 April 1886. The tree had it; the index did not. That is the exact reverse of the Mathæus duplicate settled the same week, where the index was right and the tree a year out — and it is a useful corrective to assuming an index outranks a tree.

⭐ And the anagrafe reference ties the household together ACROSS A MOVE. This entry cites Anagr. p. 305, III. 9; Magdalena Justina's baptism cites p. 305, III. 7 — same page, same family, different person number — though she was entered at house 61 in 1879 and he at house 82 in 1888. The house number moves; the anagrafe family does not. That makes the anagrafe the better key for this parish than the house number.

The household of Gregorius Defranceschi and Maria Vojnić now stands at four children buried small: the twins Maria and Anastasia (20–22 February 1875), a Maria of 1878 still unread, Magdalena Justina (1879–1952, who lived), and this Josephus.

Both birth years are now 1886 and re-declared in data/decisions.json. Nothing was deleted; ids are minted once and frozen and the join is written on both rows.

· Corrected in place: /liznjan/ · /who/josephus-defranceschi/ · /searched/

The roster carried «Mattheus Defranceschi, 1891–1891» from the MyHeritage tree and «Mathæus Defranceschi, 1890–1891» from the burial index. They are the same child.

Ližnjan «Liber Defunctorum», page 108, entry 7 of 1891, house 3: «mort. 18 Febr., sepul. 19 ejusdem, 1891» — «Mathæus Defranceschi, Mathæi et Annæ Skoko» — «nat. 25/1 90». Born 25 January 1890. The tree's «1891» is wrong and is corrected.

Nothing was deleted. Person ids are minted once and frozen — a removed id is a dead URL — so both rows stand and the join is written on each, which is what this archive did the last time it found a duplicate of its own making.

⚠ And it is not one row. Comparing all thirteen remaining tree-only Ližnjan people against the register-sourced ones of the same name and neighbouring years turns up FOUR pairs that look like the same child entered twice — once from a document, once from the tree, a year or a spelling apart:

Only the first has been read on the page. The other three are recorded as SUSPECTED and nothing has been merged, because a year's difference between a tree and an index is exactly what a register exists to settle, and guessing would put the archive back where the tree left it.

· Corrected in place: /liznjan/ · /who/mattheus-defranceschi/ · /who/math-us-defranceschi/ · /searched/

The roster carried «Tusca Defranceschi, 1870–1870», of Ližnjan, imported from the MyHeritage tree on 20 September 2026 and marked «no register has been read for this person».

⚠ But this archive's own register.json already held BOTH halves of her life, as two unjoined leads:

Same child. Same parents, in the same words. Two spellings of one name, and the archive kept them apart and took its roster entry from the tree instead.

Both registers have now been read at the page. Baptisms page 15: «9ⁿᵃ Januarii 1870», house 3, «Fusca, nata hodie», of «Matthæus Defranceschi», agricola, and «Maria n. Primus». Deaths page 54: «mortua 6. Julii, sepult. 8 ejusdem», house 3, «Fusca filia Matthæi Defranceschi q. Matthæi et Mariæ n. Primus q. Paschalis», «6 mens.», of «debilitas».

The name is FUSCA, with an F, in both entries. A hand is only legible against itself, and the burial page supplies the comparison unasked: its first entry, same page, same hand, reads «Catharina filia di Giovanni Garbin e di Fosca». Fosca/Fusca is an Istrian name in use in this parish; Tusca is not a name at all.

⚠ Her id stays tusca-defranceschi-1870 and so does /who/tusca-defranceschi/. Ids are minted once and frozen; a moved id is a dead URL. The id keeps the error the page no longer makes.

What the correction gains, beyond a letter: an exact birth date, an exact death and burial date, a house number, a cause of death, an age — and two grandfathers the archive did not have: «q. Matthæi» and «q. Paschalis», a Primus.

· Corrected in place: /liznjan/ · /who/tusca-defranceschi/ · /searched/ · /worklist/ · /register/

On 30 September this archive opened the Gologorica baptisms for 1786–87 and reported that the opening «confirms two readings already in the archive»: «Theresia Defranceschi nata Grach» and «Catharina Defranceschi nata Grach» as godmothers. The second of those has been carried ever since — on the worklist, in the Grach page's timetable and in the line of worklist 218 that sent this session to the baptisms at all.

⚠ The opening was re-read at magnification on 11 October. It gives:

There is no «Catharina Defranceschi nata Grach» on the opening, and none anywhere else this archive has read. The reading fused a Catharina from one cell with a «nata Grach» from another two rows below — and there are two Catharinas on the page to do it with, one née Jablich and one née Deprato.

⭐ What is true is better than what was claimed: Theresia née Grach stands godmother twice in six weeks, in December 1786 and January 1787 — and this is the register writing her maiden name out, which the Gračišće book refuses to do for married women.

⚠ It also removes a person from the Grah case. The claim that a second Defranceschi wife was born a Grach was part of the reason this family looked locally rooted. It was one woman, not two.

· Corrected in place: /searched/ · /worklist/ · /married-in/grach/

This archive has carried «1722–1782» on the best-documented Gologorica man of its eighteenth century, both figures off a family tree.

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

Aged sixty-four at the end of 1787 — so born about 1723 and dead five years later than published.

The death year is corrected to 1787. The birth stays at the tree's 1722 with the register's c. 1723 recorded beside it, because an implied year is not a declared one.

What this fixes, rather than breaks: the «Hellena uxor Josephi» of 4 October 1787 — his wife, seven weeks before he died; and the «filia Josephi et Hellenæ jugalium» *of April 1783 that this archive flagged and could not resolve. Both were right and the obit calendar was wrong; it is a copy made about 1800 and this is the second error found in it about the same woman.

⚠⚠ And it exposes a duplicate. A second, undocumented tree row carries «Guiseppe / Josephus / Josip Defranceschi, 1723–1787» — exactly the register's figures. The archive has very probably hung its findings on the wrong one of two tree rows for one man. No merge is made: ids are frozen, and everything written this week points at the documented row. The adjudication is on the worklist.

· Corrected in place: /person/josephus-defranceschi-1722/ · /gologorica-line/#mercatordeath1787 · /worklist/ · /searched/

Published this morning, on the strength of the 1764 and 1766 entries: a single career for a single priest. The sentence was «the ladder is 1764 no office · 1766 Reverendus Dominus · 1768 Cooperator · 1782 Parochus».

⚠ Four hours later the same book gives a canon.

> 31 Augusti 1774 · «baptizatus est ab Admodum R.ᵈᵒ D.ⁿᵒ CANONICO MATTHIA DEFRANCESCHI, de licentia Parochi hujus Loci» (ZM34K-214, M01068932).

The ladder had arranged five entries into one man's career by putting them in a line. Set out as the registers actually write them, they fall into two columns: 1764, 1766, 1772, 1774, 1781 — plain «Mathias», never an officer of this parish, always «de licentia Parochi» or «loco Parochi», and from 1774 Canonicus; against 1768 and 1782–88 — «Joannes Matthias», Cooperator then Parochus, signing in his own right.

It may still be one man — a canon of Pedena who later takes his own village's parish, needing leave only while Bancich holds it to 1782. ⚠ It may be two. The 1798 death entry abbreviates the parish priest to «MATH. de Franceschi», so the forename test does not settle it either.

What is withdrawn is the assertion of one career. The two columns and the open question replace it.

---

✔✔ RESOLVED THE SAME DAY, AND THE REGISTER DID IT.

> 24 Februarij 1782 · «Ego JOANNES MATHIAS FRANCESCHI, CANONICUS et PAROCHUS GOLLOGORITIÆ» (ZM34K-214, M01068941, in his own hand.)

The two columns are one man. He writes both titles in one phrase and uses the long forename to do it.

The withdrawal stands — the ladder was wrong as published, because it contained no canonry at all and made him Cooperator in 1768 and Parochus in 1782 with fourteen blank years between. What replaces it is the sequence the register itself narrates: priest with no office here 1764 · Reverendus Dominus 1766 · priest «loco Parochi» 1772 · Canonicus from 1774 · deputising for an ailing Bancich from December 1781 · Canonicus et Parochus from February 1782. Told in full here.

· Corrected in place: /gologorica-line/#supana1767 · /gologorica-line/#canon1774 · /person/joannes-mathias-franceschi-gologorica/ · /searched/

Published this morning, on finishing the Gračišće sweep: that the back of the 1667–1745 volume holds «two confirmation rolls nobody had counted», presented as a structural surprise — «folios 165–175 are not baptisms».

⚠ The archive knew. It had written it down twice, nine days earlier. The film's own printed title card, transcribed here on 30 September:

> «ŽUPA: GRAČIŠĆE · MATIČNA KNJIGA ROĐENIH 1667-1745 · MATIČNA KNJIGA KRIZMANIH 1690-1694 · SIGNATURA: 3.1»

— krizmani, the confirmed. And the frame map on the Gologorica line page has carried the row «Rođeni · Krizmani · 1667–1745 · 1690–1694» ever since. So has the Record Atlas's entry for DAPA, which lists «Gracisce baptisms 1667-1745 with confirmations 1690-1694».

⭐ What survives, because it is the part that was actually done. The rolls had never been read. Reading them, line by line on every leaf, produced the finding: no Defranceschi child and no Defranceschi sponsor in either, and that nil does not share the weakness the 1732 decree exposed in the baptism half. The evidence is as strong as it was. Only the claim of novelty was false.

⚠ THIS IS THE SEVENTH TIME IN A FORTNIGHT, AND THE PATTERN IS NOW THE FINDING. Six earlier self-duplications are already recorded — a woman's origin, the priests' origins, a house number, a page, a Croatian phrase, a whole argument — each republished as new when the archive already held it. This one is different in a way worth separating out: the others were things the archive had found and forgotten. This was a thing the archive had catalogued and never examined. A catalogue entry is the easiest place in the world not to look, because reading it feels like knowing it. Added to the method as a rule of its own: before claiming any part of a source is unexamined, grep the archive for the source's own title card, not only for the surname.

· Corrected in place: /gologorica-line/ · /worklist/ · /searched/

The 5 May 1744 entry — the fifth and last of this archive's Defranceschi occurrences in the 1667–1745 book — was published as «D.nus Joannes Petrus Defranceschi et Mattia uxor supradicti Joannis Grach».

⚠ The word is «Sup.ⁿⁱ» — Suppani, of the headman. doc, M01070111, folio 160, read at six times.

⭐⭐ And the difference is not pedantry — the published reading contained an impossibility nobody noticed. «Supradicti», the aforesaid, points back at the only Joannes Grach named earlier in the entry: «Præsbiter Joañes Grach», who performs the baptism. It made the godmother the wife of the officiating priest. The true reading does the opposite work: it distinguishes two men of one name — Presbiter Joannes Grach, who baptises, and Suppanus Joannes Grach, the village headman, whose wife Mattia stands.

Settled by the scribe himself, on the same leaf. Three weeks earlier, 11 April 1744, the same hand writes the man out in full: «Patrini fuerunt Sup.ⁿᵘˢ Joannes Grach et Mattia uxor eius». The archive's rule is that a hand is only legible against itself; here it did not have to look further than the facing column.

⭐⭐ What it gives back is better than what it takes away. The entry now reads as an exact pair:

> «Patrini fuerunt D.nus Joannes Petrus Defranceschi et Mattia uxor Suppani Joannis Grach»

— set beside 26 June 1743, «Joanna uxor Suppani Jo: Baptistæ Defranceschi de Golagoriza». Two headmen's households at one font, one from each village, standing for the children of the judge of Gračišće. That is the rota this archive has just described, and the misread word was hiding half of it.

Also recovered: the child has a name. This archive had «a daughter» of Ignatius Martinich and Lucia. The register says «Joannam». ⚠ Note without asserting: their 1740 daughter is «Joanna Petronila». Two living daughters of one name is unusual and the ordinary explanation is that the first died — but no burial has been found, and this archive does not kill a child to tidy a list. Left open.

· Corrected in place: /direct-line/ · /gologorica-line/ · /person/ · /worklist/

The direct-line page transcribed the 19 March 1734 baptism as «Presbiter Philippus de Franceschi Baptisavit Josephum Antonium filium legitimum Ignatij Marinich et Uxoris Luciae» — and glossed it «lawful son of Ignatius Marinich».

⚠ The word is Martinich. Magnified nine times from M01070088, folio 138: between the r and the i stands a looped ascender with a crossbar. «Marinich» has no letter to put there. (doc.)

⭐⭐ Why one letter was worth the crop. The archive already holds four Defranceschi appearances for one Gračišće couple — Ignatius Martinich and Lucia — in 1744, 1747, 1750 (and Helena in 1752). Spelled «Marinich», the 1734 entry is an unrelated baptism a priest of the name happened to perform. Spelled «Martinich», it is the same couple, ten years earlier, and the tie becomes sixteen years and four children — opening with a Defranceschi priest at the altar rather than a Defranceschi neighbour at the font.

⚠ NOT touched, and worth saying so because the string is the same. «Joh. Marinich, Canonicus Gallignanensis», who buries Sancta and Catharina Defranceschi in 1792, is a different man in a different register. That reading has not been checked here and has not been changed. A global replace on this surname would have corrupted it, which is the argument for fixing transcriptions one entry at a time.

· Corrected in place: /direct-line/ · /gologorica-line/

Published this morning, in the block on the 1732 visitation decree: that the signature «Fran.cus Augustinus Corsi Vic.us Gen.lis, et Præpositus» names the same cathedral office of Pedena that baptised Count Damian de Franceschis' son in 1720 under «Marino Iosepho Bellaz Præposito, nec non Vicario Cathedralis Ecclesiæ Petinensis» — reasoned, and flagged as inf, from the fact that «Præpositus paired with Vicarius is not a form two different posts share».

⚠ It is. The register settles it four frames later, and the archive did not have to wait for an outside source. On M01070086, folio 137, Corsi baptises in his own person and gives his style in full:

> «Die 3 Februarij 1733. Nicolaum filium legitimum Judicis Joanis, et Joannæ Uxoris jugalium Jurich Salamon Baptizavi de Licentia Parochi Fran.ᶜᵘˢ Augustinus Corsi Vic.ᵘˢ Gen.ˡⁱˢ Petenensis et Præpositus Pisini»

Vicar General of Pedena and Provost of PISINO — Pazin. Bellaz was Provost and Vicar of the Cathedral Church of Pedena. Both men are dignitaries of the same small diocese; they do not hold the same post, and Corsi's provostship is at Pazin, fifteen kilometres away and a different church.

What survives, and it is the part that mattered. Two named holders of Pedena-diocesan office, twelve years apart, both physically at Gallignana and both leaving their hand in this register — and Corsi is still the handle on a visitation file. What changes is where to ask: not only the Pedena cathedral chapter, but the provostship of Pazin, whose papers are a DAPA matter rather than a Trieste one. Worklist 212 redirected.

Why it happened, since the cause is reusable. The archive's own rule is that a hand is only legible against itself — and the rule's other half, less often said, is that a title is only legible against the same man writing it in full. The decree abbreviates: «Vic.us Gen.lis, et Præpositus», with no place attached to Præpositus. An abbreviation was read as if it were a construction, which is exactly the failure the evidence rules already name for rank and trade — and the second construction was four frames ahead in a sweep already under way. The fix: do not publish an office off an abbreviated signature until the sweep that would give the second construction has passed it.

· Corrected in place: /gologorica-line/ · /direct-line/ · /worklist/ · /gallery/

The trees page described this family's drawing as: «Now drawn twice: as a pedigree with seventeen ancestors and eleven walls, and as a channel of places — Gologorica, Gračišće, Crikvenica, Klenovica, Pfullingen, Senj, Johannesburg, Brisbane.»

The channel is defined in the data and is not drawn. It has no renderer: five pages pass the shared kit's Spine a prop it does not take, and the data would not fit the prop it does take — the kit draws a ladder of couples, the data describes ground covered.

On /tree/ the section «The line as ground covered · Eight generations, and the seven places that hold them» has a label, a heading and an empty body. The same is true on /seget/, /corsica/, /imotski/ and /omis-line/.

⚠ What is NOT claimed: that those pages are blank. Each still draws something else — an ancestry chart, or SVG of its own — so «every one of them now drawn» holds at the level of the page. What was wrong is the word «twice», and a reader told to look for a channel of places would find a gap.

Found while checking, at David's asking, whether this week's work was reaching the charts. The row carries the five headings so the next session can find them in a minute.

· Corrected in place: /trees/ · /tree/ · /worklist/

David asked whether this week's work was reaching the family charts. Auditing that found this.

The families page summarised the evidence for the Carnian family as: «A De Franceschi of Mione on the 1679 headmen's list … and an unbroken register descent at Gračišće from 1787.»

There is no register entry at Gračišće in 1787. Joannes Baptista's birth of 13 August 1787 comes from the Status Animarum of house 5 — a compilation — and his baptism is not in the Gračišće baptism book, which has now been read: the August 1787 entries are 30 July (Fernasar), one cancelled «rescissa», and 23 August (Vojscovich).

And the wider sweep makes the wording worse, not better: across 1746–1808 the surname never appears as a father at Gračišće at all, and its six appearances are all as godparents, five labelled «of Gologorica».

What is corrected and what is not. The descent is not in doubt — it rests on the Status Animarum of house 5, on the registers after 1808, and on the 1863 baptism this archive holds a certified copy of. What was wrong is where its earliest evidence sits, and «unbroken register descent from 1787» told a reader there was a baptism entry to go and look at. There is not.

Also done in the same pass: five households added to the household data — the direct line's own house 5 at Gračišće, and four at Gologorica — because that file held 132 households and not one from either of this family's origin parishes.

· Corrected in place: /families/ · /direct-line/ · /households/

This afternoon this archive published that the epithet «Gollagoriciensis» on a 1750 Gračišće godfather reversed its understanding — «This archive has treated Gračišće as where the direct line comes from and Gologorica as a second home. The register of 1750 points the other way.»

It had already been written down. The archive's own note on Count Damian de Franceschis says:

> «The whole Gallignana baptism book of 1667–1745 has now been read leaf by leaf, and the surname occurs five times in it … and three times a Defranceschi of Gologorica standing godparent, in 1733, 1743 and 1744. The count is the only one of the five who is not from Gologorica.»

So the pattern — Defranceschi appearing at Gračišće as godparents from Gologorica — was established before today, in the book before the one swept today.

What survives, and it is still worth having: the six appearances of 1747–1783 are new, their frames are new, and they show the pattern continuing unbroken into the next volume — with five more epithets and the Martinich judges asked five times in forty years. The finding is a continuation, not a reversal, and the page now says so.

This is the sixth self-duplication in three days. The check that would have caught it is not a name or a frame but a claim — before writing «this changes what the archive thought», grep for what the archive thought.

---

⚠⚠ AND THIS CORRECTION WAS ITSELF PUBLISHED BEFORE IT WAS TRUE. It said «the page now says so». It did not. The script meant to amend the paragraph matched on THIS ARCHIVE HAS TREATED GRA against text reading This archive has treated Gračišće — a case mismatch. The loop ran, matched nothing, changed nothing and exited cleanly; the commit went out naming two files and carrying one. Caught by reading git show --stat, not by the gate, which cannot see a claim that is merely false. Fixed in the following commit, and a method rule added: assert the new text is in the file after writing it, and never let an anchor miss pass silently.

· Corrected in place: /gologorica-line/

On 6 October this archive wrote that the death book's own list of outsiders made Krbune «the place Gologorica's incomers came from — more often, across ninety-six years, than anywhere else named here», and redirected a work-list row on that basis.

The claim rested on two entries out of six.

The six are the «Extranei» column of the death register — people who died at Gologorica and did not belong to it. That is a real list and a good find. It is not a sample of migration: it is a sample of who happened to die here while a stranger, and six of anything will not carry a superlative.

The census keeps the actual traffic, and it says something else. Thirty openings of the «Locus ex quo vel ad quem» column (M01069319–M01069366) name Paz five times (houses 5, 7, 13, 24, 28), Gradinje three (24, 27, 28), Novaki three, Draguć twice, Gračišće twice, Pazin twice — and Krbune not once that can be read with confidence.

What is withdrawn and what is not. The sentence calling Krbune the commonest origin is withdrawn. Krbune is not wrong — the two «ex Cherbune» entries stand, the waypoint is still unused, and eighty-nine openings of the column are unread. What is gone is the reason given for putting it first; on the evidence in hand that place is Paz.

The rule it earns: a list of six is not a sample, and a superlative needs a denominator. Before writing «more often than anywhere else», name the set being counted and say how big it is.

· Corrected in place: /gologorica-line/ · /worklist/

This archive published the origins of the eight parish priests of Gologorica from the 1826 census, and named seven of them. The third, Blasius Zwölber, was given as «e Zuckfeld» — left in italics and unexplained, because «Zuckfeld» is not a place.

> M01069319, at nine times magnification: «e Gurkfeld»

Gurkfeld is the German name of KRŠKO, in Lower Carniola, on the river Krka — the Gurk — which is where the name comes from.

The reading was tested the way this archive's own rule requires, and the test was free. «A hand is only legible against itself: find the same scribe writing both candidate names elsewhere in the same book.» Two lines below, in the same cell and the same hand, stands «e Zirklach» — a known Z, flat-topped with a low loop. The letter in question is a tall G with a full descending loop. They are not the same letter, and nothing had to be resolved by staring.

Why one word is worth a row here. It completes the set: all eight priests now come from identifiable places — Vrbnik and Omišalj on Krk, Krško and Cerklje in Carniola, Pazin, Novaki, Lindar. An unidentifiable origin in a table of identifiable ones is a silent invitation to skip it, and it had been skipped.

· Corrected in place: /gologorica-line/

Yesterday this archive wrote that «eight Defranceschi households open the census» and set them out as a table. Reading all eight pages in full for a different purpose showed that seven of them contain no household.

The error was reading a page heading as a family. [The printed column is «Nrus. Inquilinorum juxta eorum familias» — tenants — and the book is a list of houses, each named by whoever answered for it. A Defranceschi at the top of a page means the house was his, not that his family lived in it.

What survives, and it is the better claim. The family does head eight of the book's first thirteen pages, ahead of every «rusticus» in the village, and other families are living in their houses. That is stronger evidence of standing than «eight households» ever was — it is evidence of property. The table is corrected in place rather than removed, because the headings themselves are the finding.

· Corrected in place: /gologorica-line/

Published this afternoon and corrected the same day. This archive wrote that in the 1826 census's paired house numbers «the TOP figure is the live one», on two grounds: the Udovicich cell has an inked «36» above a struck «64», and the upper figures climb in order down the book.

The ordering argument was sound. The word «top» was not.

> M01069365, the Petrus Rabar household: «34» faint · «37» faint · «38» inked

Three figures, and the live one is at the bottom. So the rule is not where a number sits in the cell — it is the ink and the strike: the live number is the one written in ink and not struck through, and the book's own order is the check on it, not the definition.

Nothing built on the claim moves. The live sequence still climbs (… 36, 38, 39, 40 …); the second system is still not the labelled «vetus/novus»; and the reading that Leonardus Defranceschi's house was 6 is still left unmade for want of a household whose two numbers are both independently known. Only the shortcut is withdrawn — and a shortcut that works on one example is exactly the kind of rule this archive should expect to lose on the second.

· Corrected in place: /gologorica-line/

This archive published 21 July 1761 and recorded that the tree said 1762, keeping its own figure because the archive's came «off the register». The 1826 census gives the date outright.

> «Jacobus Defranceschi · civilitas · 21 | Julii | 1762» (M01069323)

The row's own reasoning was right and pointed the wrong way. It argued that a one-year gap of this shape is «one year of arithmetic apart, not one of evidence» — because a record giving «aged N» yields a birth year that depends on whether the birthday had passed. That is exactly what happened, and the arithmetic was this archive's own. The register gave an age; the census gives a date. It agrees with the tree, and it agrees with the «88» on his death entry of 8 December 1850.

The person id stays jacobus-de-franceschi-1761 — ids are minted once and frozen; a moved id is a dead URL. Only the dates move. His death day remains corrected to 8 December from the register, against the unsourced 3 December.

The lesson is cheap and worth having: when this archive and a tree differ by exactly one year on a date derived from an age, the tree is as likely to be holding the original date as to be wrong — and the thing to do is find a record that states a date.

· Corrected in place: Jacobus de Franceschi · /gologorica-line/

This archive's errand for the Pićan registers described them as «MKR / MKU Pićan (Pedena), diocese of Poreč-Pula». For the years that errand asks about — the 1880s to the 1910s — that is the wrong diocese.

> «Vidi in visitatione canonica die 7. Julii 1876. G. Dobrila, Ep.s» (deaths p. 128, M01069293)

Juraj Dobrila was Bishop of Trieste and Capodistria from 29 May 1875 until his death in January 1882 — he had left Poreč–Pula in 1875. A bishop makes a canonical visitation of his own diocese, so his signature in this book says plainly which diocese it belonged to.

The chain behind it was already on this site and had simply never been joined up. Gologorica was a parish of the diocese of Pićan until its suppression on 20 August 1788 — Orbanić's edition of the 1788 spiritual visitation covers it in the diocese's last year. Pićan's territory went to Gradisca, and when Trieste was restored on 12 September 1791 the former Pićan territory stayed in it. Central Istria did not pass to Poreč–Pula until 17 October 1977.

What was wrong and what was merely incomplete. The errand itself went to DAPA, which does hold the parish books, so no letter went astray. But the description was a statement about the past written from the present, and anything built on it — a request for marriage dispensations, a visitation protocol, an ordination file — would have gone to Poreč or Pula for records that are at Trieste. The ref is corrected and a worklist row now carries the Trieste diocesan archive.

The rule it earns: name a diocese with the year, the way this archive already names a house number with its numbering. A see is not a fixed attribute of a village — Gologorica has been in four in two hundred years.

· Corrected in place: /errands/ · /gologorica-line/

was Every box in the family figure carries a band saying how the person is known: the archive's own line · a register read here · a FamilySearch index entry · the family tree, unverified. The rule that chose it read «if this row is flagged as the compiler's kin, say «the archive's own line»» — and checked that before it checked where the fact came from.

now mine is a statement about kinship and it was being printed where the reader looks for provenance. Ninety-four rows carry both that flag and a MyHeritage identifier, and every one of them was claiming the strongest band on the page — including every person the Australian and Crikvenica branches brought in on 15 and 16 September, whose only source is David's family tree. The evidence wins now. The drawn descent keeps «the archive's own line»; a row read off a register page — the Gologorica and Gračišće work — says a register read here, which twelve of them were not getting either; an ark says index; a tree identifier says the family tree, unverified; and kinship with no other source falls through to «line» last. Thirty-two people now read as tree-sourced and fifteen as register-read, where before both were hidden inside twenty-four «own line».

how David, looking at Antony De Franceski's page: «petrica was a woman and has children, you arent reading myheritage gedcom?» She is in the archive — her page draws her husband and both daughters — but on her brother's page she was a bare name in a box, and the box was claiming a provenance she does not have. A sibling box now says when that person heads a household of their own, so a sister with two daughters stops looking like the end of a line.

· Corrected in place: /who/ — the band under every name in the family figure

was The three maps were merged with their own palettes carried over intact, and for a day that was harmless because only one legend showed at a time. Layering them broke it in the same stroke that made them useful.

now Measured in CIELAB, eight pairs of colours across the three legends were closer than ΔE 20 — the point at which two dots stop being reliably tellable apart — and two pairs were effectively one colour: · «Every volume read» and «Both, at one place» were the same hex value, ΔE 0.0. · «Researched, books unopened» and «Where they went» were ΔE 1.7 apart. · «Istria, where the name is» sat 12 from one and 19 from another. Re-cut as three families that cannot be confused for one another: the shelf is a warm ramp — red for catalogued and never opened, orange for researched, amber for a volume read, green for finished, grey for not in any catalogue; our people are violet through pink; the ground is blue. No pair is now closer than ΔE 21, and the only pairs near that are inside the people family, which is one question's sub-categories. And hue is not the only channel any more. A marker's ring says which layer it is being coloured from — thin white for the shelf, deep violet for our people, navy for the ground — because a reader looking at one dot should not have to work out which of three legends it belongs to. Red also does work the old sand did not. Twelve hundred of the 1,235 shelf places are catalogued and never opened, and the map now looks like that.

how David, on Pićan: «it looks like the colour is «every volume read» or is the colour «researched, books unopened»? colours are way too close to one another. if not opened it should be red». Pićan was «researched, books unopened» and sat beside a gold marker it could not be told from.

· Corrected in place: /map/

was The page opened: «Anton Mario De Franceski, born 21 August 1944 at Reutlingen — Zdravko and Ruža's only child, and the four-year-old who sailed with them.» Below it, in the same column, the archive printed «No family is drawn for this name, and no record stands behind it» — under four sentences naming his parents, his father's date of birth and the file they are all three on.

now He had a sister, and this archive found her the same afternoon it wrote «only child». Petrica Defranceski was born at Horsham nine years after the family landed, and she is the reason the whole Australian branch came into the archive at all — David asked why she was not here. The second sentence was worse than wrong, it was structural. households.json is built from registers, and this family's modern generations have never been in a register this archive can reach, so the mini-tree had nothing to draw from — for Antony, for Zdravko, for David's own father, for David. Nine households have been added from the family's own tree and the documents already held, marked as tree-derived and not register-read. Antony's page now draws his parents, his sister, his wife and his two children. And thirty internal links across the person pages were printing as literal square brackets — including the one on this page — because the renderer that page keeps has never understood a link. It does now.

how David: «this doesnt make sense. if he is Zdravko's son then you should be able to draw this tree? along with his sister, and kids». Both faults were visible on one screen and neither had been noticed by the archive that wrote them.

· Corrected in place: /who/antony-de-franceski/

was 199 roster rows carried the mine flag, which drives the «★ My line» filter on the people page and decides who is counted as this family rather than as somebody else of the name. Nothing ever tested it. Two of the 199 were Mario Defranceschi of Brisbane, corrected earlier today.

now A test, 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 actually drawn. A shared surname on a gravestone is not a path. Twelve more rows fail it, and they are all Find a Grave memorials with no join to anything: Joachim Bruno and Gerda at Millbrook, Conchita in the Bronx, Joseph and Basilio at East Farmingdale, John A. and Bruno at Middle Village, Edward «Dado» at Hartsdale — the American wave — and Adelia at Lilydale, Corrado at Fawkner and Attilio at Parkville, which the diaspora page had already filed under the Italian post-war assisted migration, and which it says in as many words is not the same movement as this family's. The flag comes off all twelve. They stay in the archive, where they belong; they are no longer counted as kin. 185 rows keep it, every one of them either on the direct line, carrying a named relationship, or standing in a household this archive has drawn.

how Written into the work list this evening as a consequence of the Mario error, and run the same night rather than left.

· Corrected in place: /people/ — the «★ My line» filter · /who/ · the map

was The archive cites «MyHeritage family tree export, July 2018 — a GEDCOM of 8,286 individuals, 254 of them carrying this surname» as its family-file source, and 513 roster rows carry an identifier from it. A reconciliation note of 7 September 2026 already recorded that «the 2018 GEDCOM is superseded and must not be used again» — the live tree holds 15,643 people — and read 459 individuals of the surname out of the live tree. That reading was never imported. The roster was left as it was.

now David found it by looking at his own tree: «Anton has a sister in Horsham. My heritage tree has all that info. Why don't you?» Because the roster had never been rebuilt from the live tree. Read out of it on 15 September, tree 4: · Petrica Defranceski, Zdravko and Ruža's daughter — the first of this line born in Australia, at Horsham — and her husband Michael Baviello, their daughters Antonia and Carla, and two grandchildren. · Anton's own household: his wife Marion Bain, his son Dean, and his daughter Lisa, 1968–2015, born at Adelaide — whom no search of this surname would ever have found, because the tree spells her Defraneski, missing the c. · Zdravko's three other siblings, all born at Crikvenica: Josip Petar «Pepo» 1911–1980, who never left; Petar Vid «Pero» 1913, the Petar of the Pfullingen file, died in Germany; and Anka Nera, 1915–1944, dead at Belgrade aged twenty-nine — about whom this archive knows nothing at all. Fifteen people added. The whole live tree still has to be reconciled, and this is one branch of it.

how David, looking at the tree on his phone. The archive had the 2018 file's people and had mistaken that for having the family.

· Corrected in place: everywhere the roster is read — /who/, /people/, the map, the atlas

was Two roster rows — «Mario De Franceschi, 15 Sep 1938 – 29 Jan 2013» off a Find a Grave memorial at Centenary Memorial Gardens, and «Mario Defranceschi, 1938–2013» out of MyHeritage — were both flagged mine, this archive's marker for the compiler's own family. Brisbane's marker on the map showed him and nobody else. Asked why his own household was not there, I told David the map was showing «your father Mario». It was not.

now David's father is Ivan Anton Defranceski, 1951–2026, generation 7 of the direct line: Senj → Johannesburg → Brisbane. The direct line has no Mario in it anywhere. In David's live tree, Mario Defranceschi 1938–2013's parents are «QLD Defranceschi (Deceased)» and «Unknown Defranceschi» — placeholder nodes, a stub for an unconnected Queensland family of the name. And the diaspora page had already filed «Mario at Brisbane, 1938–2013» with the Italian post-war assisted migration of the 1950s and 60s, which it says in as many words is not the same movement as the Croatian displaced persons of 1949. The flag is removed and he is filed under other families of the name. Two errors, and the second is mine: the archive mis-flagged him, and I read the flag off the screen and told David it was his father rather than checking the direct line, which was one file away.

how Found while pulling the Australian branch out of the live tree, by asking who Mario's parents were.

· Corrected in place: /who/ · /graves/ · the map, at Brisbane

was The map carried four colours and the third was «Something here read» — gold. Twenty-two markers were gold. The page said the scale meant «how far we have got» with the shelf, and the panel on a gold marker said «Something here has been read — the search register says what. This archive does not yet track WHICH of these volumes.» Read together, that says: we have opened books here and not written down which ones.

now Twenty of the twenty-two had never had a single register volume opened. They were gold because the search register NAMES them — and what it names them for is literature. Žminj is the clean example: the two rows that flag it are «Libri žminjski, libar drugi» (Žminj 2008), a local history sent by David, and Darovec's 2021 migration paper, which lists Žminj among the towns the Carni settled at. Both were read. Neither is a register. Only Mione (15 volumes) and Ovaro (1) had a book of their own opened, and both are Antenati, the one provider whose catalogue carries a per-volume flag. Reading a book about a parish and reading the parish's own registers are two different acts, and a scale whose legend says «how far we have got with the shelf» must not paint the first in the colour of the second. The two are now separate colours: teal — researched, its own books unopened, and gold — a volume of its own registers read. The map loses twenty gold markers and gains a state it did not have a word for.

how Found by David, looking at Žminj: «it's on the map as something here read but then none of the volumes are marked as part read or read». The per-volume marks added an hour earlier are what made the contradiction visible — the place said one thing and every book under it said another.

· Corrected in place: /map/ · the marker colours and the panel

was The Atlas carried 52 hand-written pins and had done since it was built. The dek said «every place this archive can put a document in», which was not true: the roster names 266 place strings attaching 1,440 people to the ground.

now Rebuilt from the roster, the map holds 154 places and 955 people. Three things were wrong rather than merely missing. · One town was being drawn several times under different names. Almissa and Omiš are the same place; so are Fiume, Vitipolis and Tarsatica, which is why the first rebuild put Rijeka on the map three times. Historic and Italian forms are now folded onto the town they name. · This archive holds two different coordinates for Gračišće, seven kilometres apart — one in places.json, one in atlas.json. Neither was flagged because nothing had ever compared them. The heavier pin keeps the ground and the builder now refuses to draw one name twice. · Seventy-three entries name something that is not a town. Thirteen say «Österreich», five «Austria», four «Italy»; and four each say «Santa Maria Assunta», «Immacolata Concezione» and «Assunzione di Maria Santissima» — church dedications written into a place field. Those are left off the map and are a data error to be cleaned, not a place to be pinned.

how Found by building the thing properly rather than by checking it. Nothing compares a gazetteer against itself until something has to draw both.

· Corrected in place: /map/

was Published hours earlier, on finding FamilySearch's Udine images restricted: «Settling Ovaro now means a Family History Center, an affiliate library, or the Archivio di Stato di Udine itself — not an afternoon at a screen.»

now The Archivio di Stato di Udine has published them itself, on Antenati, free and open. Ovaro has 271 registers there and fifty-nine of them are Mione's, catalogued under its own name. The first correction was right about FamilySearch and wrong about the world. One vendor's gate was mistaken for the state of the record. The reason it was missed is worth keeping: Antenati was searched by name, which returns nothing for Tolmezzo, because only part of the holding is name-indexed. The portal says so on the results page. The registers were one form away the whole time.

how Found by retrying a source this archive had twice written off — once as «403s both browsers», once as «needs a reading room» — and by reading the failure message instead of accepting it.

· Corrected in place: /parish-books/

was Published this afternoon: that Mione's fifty-nine registers are open on Antenati and that «this archive has been saying for months that they were not online».

now The Carnia page already said it, in as many words: «The Archivio di Stato di Udine has digitised the Napoleonic civil registers of the Comune di Mione — fifty-nine volumes… and publishes them openly on the Antenati portal.» It had already read the 1809 and 1810 birth books and published four households out of them. What was actually new this afternoon was narrower and should have been stated as such: the 1806 volume had not been read, and it holds a fifth household and a generation above it. The error was mine and it was the ordinary one — I searched the web before I searched the archive. A site of twelve hundred pages will tell you what it already knows if you ask it first.

how Caught within the hour, by grepping this archive's own data before writing up the register act.

· Corrected in place: /searched/

was Published this evening, as the closing note on the Ovaro marriage allegati: «Mione and Muina are both done, and both are closed.»

now Mione is not closed. The Antenati register browser returns its results ten to a page, and I read page one. Page one carries two Mione files. Pages two and three carry ten more — 1808, 1809 (two files), 1810 (two), 1811 (two), 1812, 1813, 1814 and 1815. Only the 1806 file had been read. Ovaro itself has eight further allegati on those same pages, and none of them had been counted either. The error was the shape of the tool, not the archive: a paginated result set answered as though it were the whole set. The fix is the obvious one — page to the end before saying «all».

how Caught the same evening, going back for the remaining village files and finding the count did not match.

· Corrected in place: /errands/

was The Muina chain said of Antonio de Franceschi and Maria née Corva: «Farmers of Muina. Both alive in 1808: Antonio gave his written consent to his son's marriage on 25 October 1807, three months before the wedding.»

now Antonio was dead by the wedding day. The marriage allegati of Mione 1808 — the loose documents the couple actually produced — style the groom «Osvaldo del fu Antonio de Franceschi» and take a separate consent from «Donna Maria nata Corva, vedova del fu Antonio de Franceschi». Under the Napoleonic Code a mother's consent was required only where the father had died, so the existence of that second act is itself the proof. The error came from the marriage act's phrase «figlio di Antonio de Franceschi, e di Maria nata Corva, pure agricoltori» — «also farmers». That names a trade, not a living man, and it was read as though the present tense covered both parents. The correction yields a fact the archive did not have. Antonio's written consent is dated 25 October 1807 and his widow's consent 28 January 1808: he died inside those fourteen weeks, bracketed by two documents, neither of which is a death record.

how Caught by reading the allegati themselves instead of the register's summary of them — and only because the file had been wrongly counted as read.

· Corrected in place: /carnia/#muina

was Published this afternoon, from the 1809 marriage allegati: that Orsola de Franceschi was bundled with the baptismal extract of «Giovanni Pietro Antonio Cramaro», born 28 April 1758, «almost certainly the groom», and that «the name Cramaro is worth a second look» because the cramars were the Carnian pedlars and «a De Franceschi daughter marries a man whose family name is that road».

now The pairing was right. The name was not, and neither was the year. The marriage act itself — Mione, Matrimoni 1809, act N. 1, read this evening — gives the groom as «Erman de Franceschi Gio: Pietro», aged fifty-six, son of the late Pietro Erman and Maddalena née de Franceschi, and reads his birth act as 28 April 1752. He signs himself «Gio: Pietro Erman de Franceschi». The priest's extract had said «figlio di Pietro cramaro del villaggio di Mione» — Pietro the cramar, the trade, lower-case, standing where a byname goes in these registers. This archive read a trade as a surname and then built a paragraph on it. The birth year was a second, smaller error of the same kind: the last digit had been cropped out of the magnified image and was guessed at from context. The truth is better than the mistake. The family Orsola married into is Erman — and «Zuane d'Erman» is one of the three men in the 1679 list who went from Mione to Istria with Zuane De Franceschi. The two families were still marrying each other a hundred and thirty years later, and the groom's own mother was a de Franceschi already.

how Caught by going to the marriage register instead of stopping at the allegati — which is the rule this archive should have been following: read the loose papers to find the marriage, read the register to believe it.

· Corrected in place: /carnia/#mione1809

was This archive has carried Appolonia Salante, daughter of Antonio Salante and Speranza née Micoli, for months — in the Muina chain, the 1806 households table, the plate caption for the marriage act, the dossiers and the open questions. And this morning, faced with the notary writing Galante twice, it doubled down: «This archive keeps Salante, because that is what the parish book and the register say, and the register is what the family was entered under.»

now The register does not say Salante. It says Galante, twice. Mione, Matrimoni 1808, act N. 1, read this evening: «…ed Appolonia Galante villica di professione, dell'età d'anni venticinque e mezzo, figlia di Antonio Galante e di Speranza nata Micoli» — in the same clear looped G the notary used. The count is now four hands to two. For Galante: the notary Giacomo Micoli in both of his consent acts of 1808, the marriage register in both of its mentions, and Pietro Galante signing his own name at Mione in 1813 in a plain lower-case g. For Salante: the parish priest's baptismal extract, and the clerk who docketed the allegati file. The name has been changed to Galante throughout — the Muina chain, the households table, the plate caption, the dossiers and the questions. The priest's Salante is kept on the record as what he wrote, because he was copying from the baptismal book and that is where the variant lives.

how Caught by reading the 1808 marriage register instead of quoting it from memory. The claim that «the register says Salante» was never checked against the register — it was inherited from an earlier transcription of the same act and then asserted as evidence against fresh, better readings.

· Corrected in place: /carnia/#muina

was Published 14 September 2026: that the 1815 death act's «Anna de Franceschi … maritata col suddetto Paolo de Franceschi» admits two readings — that she was born a de Franceschi, as the 1808 act's «Giovanna pure de Franceschi» shows is possible at Mione, or that the clerk entered a married woman under her husband's surname — and that this archive would not choose between them without a second document.

now The second document turned up the next day, and it chooses. The Mione banns of 4 June 1815 — five months before she died — publish the marriage of her son and name her «Anna nata Brunis». She married into the surname. The endogamy reading is wrong for this couple. What survives: the 1808 act still says «pure de Franceschi» in as many words, so one De Franceschi–De Franceschi marriage at Mione stands. Not two. The maiden surname is read as Brunis; Brencis, already documented in this village, cannot be excluded at this resolution.

· Corrected in place: /carnia/#mionepaolo

was Published on the Research Map's first build: that thirty-four of the forty Kvarner parishes carry only a browse link and no volume list at all, that the rebuild was what made this visible, and that those thirty-four were first in the queue for enumeration. The page's coverage figure was given as 111 of 1,196 places and 590 volumes.

now They had all been walked, and the fault was in my own builder. Kvarner keeps its volume lists under a field called films — the same as Dalmatia — and carries a stale books title-list on six of its forty parishes. The build script read books for Kvarner and films for Dalmatia, so thirty-four parishes with full inventories came through as empty. Corrected figures: 140 of 1,196 places walked, and 777 volumes inventoried — not 111 and 590. Kvarner's 255 volumes are on the map, with links that open each book at its first image. This was caught within the hour, and only because David asked me to start work on the thirty-four. The instruction was sound; the premise I had given him was not.

· Corrected in place: /map/

was The page's main paragraph carried two internal links written in markdown — Logged as a correction among them. Both rendered on the live site exactly as typed, brackets, parentheses and all.

now The archive's markup renderer has no link syntax and never had one. md.js handles bold, italic and «quotes», deliberately, so that data files stay plain. Writing a markdown link into one of those strings does not fail loudly — it just prints. Fixed on the page rather than in the shared renderer: the two links are resolved locally and put through the same base-path helper as every other internal link, so no other page inherits a syntax it was not written for. The correction that this page pointed at was unreachable for as long as the link was broken, which is the part that actually mattered.

· Corrected in place: /map/

was The research map's plan named Matricula Online as a fifth provider to be built in, «(Senj, Karlobag)».

now Matricula holds no Croatian parish at all. Its fonds index lists Germany, Austria, Poland, Serbia, Luxembourg, Bosnia-Herzegovina, Slovenia, Italy and Switzerland, and its place search returns nothing for Senj, Karlobag, Rijeka or Fiume. That line was written from memory and never checked before it went into the plan. The map has four providers, not five, and the item is struck rather than left to look like outstanding work.

· Corrected in place: /map/

was The first build of the Antenati catalogue named fifty-nine volumes for Mione and Ovaro from contiguous ark ranges, on the assumption that the portal numbers its registers a comune at a time. Series spans were written from that assumption, and one range — an_ua679984 to 679991 — was given to Ovaro's marriages outright.

now It does not number that way. Walking each ark's own catalogue page showed that an_ua667510 to 667513 are Mieli, a different comune — Morti 1807, Nati 1806, Nati 1807 — sitting immediately before Mione's marriages in the numbering. The catalogue is rebuilt from forty-six arks actually fetched and read, plus the eleven this archive has opened act by act. Every volume now carries how its identity is known — walked, read, or unconfirmed — and five arks inside Mione's range are listed as unconfirmed rather than given a year they might not have. The Mione inference survived the check. Matrimoni, allegati, indici and pubblicazioni came back exactly as predicted, 1807 missing from each. The Ovaro range did not survive and is withdrawn: only the two Ovaro volumes this archive has actually read are claimed. The page stood for under two hours. It was checked because a single probe — the very first ark fetched — came back Mieli.

· Corrected in place: /antenati/

was Published on the Carnia page and written into the Pieve di Gorto errand: that 1807 is missing from every series at Mione and at Ovaro, and that there is «no 1807 marriage volume, at Mione or at Ovaro».

now Ovaro's 1807 marriage register is on Antenati — ark an_ua679983 — and Ovaro's marriages run 1806 to 1815 with no gap at all. Its 1807 allegati survive too (an_ua679993). The claim was written from the Mione inventory and extended to Ovaro before Ovaro had been walked. Walking it took twenty-five fetches and contradicted the sentence. The correction makes the finding stronger, not weaker. A year missing from one comune while the next comune up the valley kept it is not a failure of the system in 1807 — it is one lost volume, and that is a thing an archivist can be asked about.

· Corrected in place: /carnia/#gap1807

was Published twice. First from the death act alone: «Morti 1810, act N. 1 — 28 January 1810, eight in the evening», and — reading the registrar's «un mese circa» as a fact — «born about December 1809». Then again after the birth act was found, as «dead at one month on 28 January».

now She died on 25 January 1810, eleven days old. The act opens «Comune di Mione li 25 Venticinque Gennajo mille ottocento dieci» — the number in figures and spelled out, agreeing. And the birth act of the same register series has her born Sunday 14 January 1810 at three in the morning. The «un mese circa» is the registrar's own error, not evidence. Giacomo Micoli wrote both acts eleven days apart and guessed the age of a child whose birth he had himself recorded. This archive then built a December 1809 birth on that guess, and kept the wrong death date even after finding the real birth. The lesson is the one this page keeps relearning: an age in these acts is a guess and a date is a record, and where they disagree the date wins.

· Corrected in place: /carnia/#mionenati1810

was The map offered five colours, one of them grey for «Nothing filmed here». Clicking it showed 0 of 1,187 places, every time. The legend said the state existed and the map said it did not.

now Two faults, and the second matters more than the first. The builder dropped any place with no provider block before it reached the map — which is precisely the set the grey colour was meant to show. Three places were being silently discarded; they are now on the all-providers view. And the label was a lie. Two of the three are Evangelical, and they are not unfilmed — this archive never found the Evangelical root in FamilySearch's waypoint tree, so that shelf was never walked. The colour is now «in no catalogue at all», the panel on those two markers says the fault is ours, and finding the Evangelical root is on the work list.

· Corrected in place: /map/

was Published this morning, and defended in prose: that shading by depth — one or two volumes, three to nine, ten or more — was the fix for a colour scale that had saturated. The legend read «One or two volumes», «Three to nine», «Ten or more».

now It answered a question nobody was asking. A parish with forty volumes nobody has opened and a parish with two nobody has opened are in exactly the same state, and the map was spending its only visual channel on the difference between them. The colour now says how far this archive has got: catalogued and never opened, started, read right through. Depth has not been lost — it is in the panel, in words, where «114 volumes catalogued» can sit beside «nobody has opened any of it» without either pretending to be the other. The honest picture is worse than the one it replaced: 1,219 places catalogued and untouched, 20 started, none finished, and 16 volumes of 7,159 read one at a time. That is the number the map exists to show. Third scale in two days. The first saturated, the second was pretty and wrong, and this one was David's correction, not mine.

· Corrected in place: /map/

was The archive's rule has been published since 10 September: «living relatives are listed by name only — no date of birth, no address, no place of residence, no photograph». It read as a policy being applied. It was naming exactly two living people: David and his mother.

now Every import this archive has ever run was built to find the dead. Parish registers, headstones, a reconstructed census, a surname harvest of a family tree filtered to people with death dates. Nothing in that machinery would ever have returned a living cousin, and nobody noticed, because the rule about living people looked like the reason they were absent when in fact they had never been looked for. Ten people added, name only: David's father's sister and brother, his uncle and aunt by marriage, three first cousins, his wife, and his two sons — who are generation 9 of the direct line and were not on a site about their own family. Found because David looked at his own tree beside this site and asked where everybody was. The dead had four hundred years of coverage and the living had two names.

· Corrected in place: /people/

was Written up on reading Mione's Morti for 1810: that the man who signs «Erman De Franceschi Gio: Pietro» undoes a distinction this archive had been making, that «the register does not keep the two names apart, and neither does he», and that the plainest reading is a by-name on a De Franceschi house — with a note that every page treating Erman and De Franceschi as separate families needed correcting.

now The archive had the answer already, on the same page, from a better document. His marriage act of 1809 — read and published here days earlier — gives him as «son of the late Pietro Erman and Maddalena née de Franceschi». The compound surname is simply his two parents, an Erman father and a De Franceschi mother carried together. That is ordinary Carnian practice and it needs no by-name, no re-reading of the 1679 list, and no correction of any other page. The 1810 signatures are still worth having — they are the fourth and fifth time he writes his own name — but they confirm rather than reveal. The failure was not in the reading. It was in treating a volume read end to end as if it arrived in an empty archive. Reading carefully is no protection against forgetting what is already held.

· Corrected in place: /carnia/#morti1810

was A page built this afternoon around Pazin's «not digitised» column, opening: «These are not books missing from a website — they are books that exist, on a shelf in Istria, and cannot be read from anywhere else.» The research map repeated it: «books that sit on a shelf in Istria and nowhere else».

now Pazin's column means «not digitised by Pazin». It says nothing about anybody else. Checked volume by volume against FamilySearch's films of the same parish, record type and years: 105 of the 154 are already photographed — often inside one long reel that swallows several of Pazin's shelfmarks at once. Barban's Births 1716–1750, 1750–1781 and 1781–1816 are three Pazin volumes and one FamilySearch film. Forty-nine are left — the ones where two catalogues agree nothing has been photographed. The page now leads with those, and they are the only ones worth a letter. Found by looking at the map's own panel. Svetvinčenat's volume list showed «Births 1568–1588 ↗» and «Births 1568–1588 — not digitised anywhere» on consecutive lines, which cannot both be true. The duplication I had been treating as a display problem was the archive telling me the claim was wrong.

· Corrected in place: /undigitised/

was The research map's standfirst, since it was built: «Every parish in Croatia whose registers are on film, and how far this archive has got with each one.»

now It is every parish in one collection. Croatia, Church Books, 1516–1994, plus three sources added by hand. FamilySearch publishes 3,488 collections, and Croatian places are scattered through the ones named for other countries — because a town's registers are filed by whichever state kept them, under the name that state used. Čakovec is filmed as «Csáktornya», in a Hungarian county, inside a collection named for Slovenia. Five Croatian Istrian parishes are inside an Italian one. A second Croatian church-book collection exists for the Delnice deanery and browses by film number, so no place-based walk would ever find it. The standfirst now says what the map actually is, and a page lists what it cannot see. Found because David asked whether Croatian towns might be recorded in other countries' books. They are, and the map had been quietly claiming otherwise for two days.

· Corrected in place: /map/

was A table headed Sources, hand-typed into the page: Pazin, istrianet, defranceski.com, Forebears, Acta Croatica, Antenati, FamilySearch, the Udine diocesan archive, Rijeka — ten rows, unchanged since the archive's first week.

now There are 83 sources in the register and 274 searches recorded against them. The ten were the ones that existed when the page was written, and the list never caught up while ANNO, Hungaricana, Findmypast, Trove, Matricula, 23andMe, the Statue of Liberty database, the Arolsen Archives, Mémoire des Hommes, CEMLA and eighty others were added. The table is now driven by the source register itself, grouped by kind, so it cannot fall behind again — and it points at the search register, which records every search including the ones that found nothing.

· Corrected in place: /about/

was Two days of work went into the research map, the collection walk, the undigitised list and the Antenati catalogue. None of them was in the navigation. The map could be reached only by a link inside another page.

now Five pages were orphaned — the research map, «Croatia filed abroad», «Every collection walked», «The undigitised» and the Antenati registers. All five are now in the Evidence menu beside Parish Books. Found because David asked which menu the map was in. It was in none, and the build's «nothing links to it» warning had been satisfied by a single in-page link each time — which is not the same as being findable.

· Corrected in place: /map/

was Marija Stjepanka, the second child of Josip Defranceschi and Ana Kalanj, was recorded as «8 December 1892 · Crikvenica», in the list of children born at the Crikvenica house.

now She was born on 8 December 1892 — that part is right — but she was baptised on 22 December at LEDENICE, and her parents were living at Klenovica. The entry is Ledenice baptisms, page 166, no. 83: «Godine 1892, Prosinca 8 / 22 — Marija Stjepanka — zakonita — Josip Defranceschi, klesar, i žena Ana Kalanj rodjena — R.k. — mjesto obitavanja i kućni broj: Klenovica kbroj 28, na stan». A small note above the child's name reads «u Crkvenici». Three things follow. The father is called klesar — a stonecutter — in 1892, the trade his own father Antonius held at Gračišće in 1863. The family's address is Klenovica house 28, «na stan» — in lodgings — not the Kalanj house at no. 22 but a rented room a few doors from it. And the household was still inside the Ledenice parish at the end of 1892, not yet at Crikvenica: the move must fall between December 1892 and November 1894, when Ivan is born at Crikvenica. Found by querying the FamilySearch index for the surname directly and then reading the register page.

· Corrected in place: /crikvenica-house/

was This archive has carried fourteen children of Valentino Defranceschi and Helena/Elena since 6 September 2026, and set that against «a family account of ten sons and four» — noting that the register gave only nine sons, one short.

now There are fifteen, and the missing son is Jacobus Vincentius. Under «Die 21 July 1757» in the Fiume baptisms: «Jacobum Vincentium filium legitimum Valentini, et Helenæ jugalium Defranceschi, natum 20 hujus, baptizavit D. Carolus Simon Bertoni, Patrini fuere Ill.mus D. Emanuel de Perretti, et Ill.ma D.na Elisabeth Bono» — born 20 July, baptised the next day, with a patrician couple standing as godparents. With him the sons come out at ten exactly, which is what the family has always said. (doc — image, index entry) Found by matching the index on the parents' names rather than the child's, then reading the page to confirm it.

· Corrected in place: /valentino/

was Aloysia, the thirteenth of Valentino's children, was listed as «born and buried 1757» among the five infants.

now The register gives her age, and it is four. The Fiume burial book for September 1757, entry 14: «Alovisia d’anni 4, figlia di Valentino, ed Elena Defranceschi, sepolta nella Tomba di …» — the entries around it that record infants say «di mesi 3», in months, so the unit here is deliberate. She was therefore born about 1753, not 1757, and she was a child rather than a newborn when she died. This also removes a problem nobody had noticed: Jacobus Vincentius was born on 20 July 1757, and no mother bears a second live child eight weeks later. (doc — image, index entry)

· Corrected in place: /valentino/

was The name table dated «Defranceski» to «20c–» and placed it in «South Africa, Australia, Argentina» — the form the emigrant branches settled on once they were abroad.

now It was already being written at home, in the parish register, in 1896. The last birth entered in the Crikvenica book for 1896 is Zorica, and her father is written — across a line break, with the clerk's own hyphen — as «Josip Defran=ceski Antonov». Magnified, the second line is unambiguous: c-e-s-k-i, no ch, no sch. Why it matters for a one-name study. The archive had been treating the sch-less form as something that happened to the name in transit — a spelling the Anglophone and Spanish-speaking world imposed. It is the opposite: a Croatian clerk writing a Croatian book had already dropped the Italian digraph, and the emigrants carried out a form that was waiting for them at home. The family did not lose the spelling abroad; it took its own with it. The same entry records the household as «Iz Istrije mjesto Gračišće, sad ovdje stanujući» — from Istria, the place Gračišće, now dwelling here — and gives no house number at all, which is how the register treats incomers. (doc — image, Crikvenica births, film image 696)

· Corrected in place: /your-name/

was Item 135 ended: «A Marinegrundbuchblatt gives birthplace, parents, religion, marriage and children. For a man born at Gračišće in 1833 that is a household — and the Gračišće baptisms of 1833 are on film and readable from here to match against it.»

now Gračišće is not filmed at all. It is not one of the thousand parishes in the FamilySearch Croatia collection; for that corner of Istria the collection holds only Pazin, Stari Pazin and Lupoglav, and no Gračišće volume sits under any of them. The archive already knew. The atlas entry for Gračišće carries nfilms: 0, and the September count of the Pazin inventory put 495 of 1,151 books in parishes with no film. The work-list note simply contradicted the map, and would have sent somebody looking for a reel that does not exist. What replaces it. Everything this site holds from Gračišće came from DAPA Pazin, who will search a named register for a named person over a stated period — they have done it for this family once already, for 117 kuna. Item 135 now says so.

· Corrected in place: /worklist/

was The roster row reads «Giacomo's first wife — mother of Domenico Gasparo, baptised 9 April 1731. She disappears from the book after that.»

now Carta 76 verso names Maria as Gio Battista's wife on 23 November 1731, seven months after the baptism the row is built on — «Battistina Figliola Legitima del S.r Gio: Battista de Franceschi e di Maria jugali». And Giacomo's wife in February 1728 is Elena, on carta 66, read at magnification. She does not disappear from the book. She appears in the other household. Not rewritten, only flagged. The 9 April 1731 baptism has not been read off the leaf here — it came in from an index — and this archive does not correct one unread record by citing another. The row now carries the conflict instead of hiding it, and the leaf that settles it is the one to open next.

· Corrected in place: /who/maria-de-franceschi/

was Published a few minutes earlier, closing the 1780 reading: «Where to resume: frame 84» — on the assumption that a register film runs in date order and that frame 84 therefore follows October 1780.

now Frame 84 of that volume is January 1763. Frames 81, 82 and 83 are April to October 1780; frame 84 opens «Anno vero Salutis 1763 Die prima Januarij». Both years were checked at magnification before this note was written, because the two digits are the whole of the claim: the 8 of 1780 has two closed loops; the 6 of 1763 has one loop and an open top. They are not the same character and neither reading is in doubt. The four frames are consecutive on the film — 005497885_00295 to 00298, resolved from the arks — so this is not a paging mistake. The film does not run in date order at this point. Why it nearly cost something. A reader resuming at frame 84 and expecting 1781 would have found a January, a plausible-looking year, and a page full of the right parish and the right families. Nothing on the leaf announces that seventeen years have gone backwards. It was caught only because the year looked odd enough to magnify — the same check that caught the Pazin misreading this morning, and the same check that failed to be made when the resume instruction was written. The rule, now stated on the work list: date every frame from its own heading. Never interpolate a date from a neighbour, and never infer one from a frame number.

· Corrected in place: /worklist/

was Published earlier today, on finding Antonio de Franceschi witnessing six weddings at Svetvinčenat between January and October 1773: «A man asked for six times in one year is a man the parish trusts and, more to the point, a man who lives there.»

now He was the parish bell-ringer. Sixteen frames further into the same volume, three entries of 1774–75 give him the occupation, and one puts the word in front of the name where it cannot be misparsed: «e del Campanaro Antonio Defranceschi. Testimonj» (25 September 1775). The other two read «Ant.º Defranceschi Campanaro» (27 February and 29 October 1775). The word was magnified and read letter by letter — C-a-m-p-a-n-a-r-o — before this note was written. The original sentence is not false; it is the wrong explanation. He does live there, and the priest does reach for him. But the reason is that he is already inside the building when a witness is wanted, not that six families independently chose him. A repeated name in a witness line is a job before it is an honour, and this archive reached for the flattering reading first. The same frames also stop the next overreach before it starts. «Zne Sain Campanaro» appears in February and July 1776 and January 1777, four months after Antonio last carries the title — so «the bell-ringer of Svetvinčenat» is not a claim that can be made for either man. Either the office changed hands, or the castle's several churches had more than one.

· Corrected in place: /searched/

was The research map has carried 1,245 placed pins with no visible distinction between a coordinate that is the place's own and one borrowed from somewhere larger. Four of them were the latter and said nothing about it.

now Found by running somebody else's test. The D'Arcy one-name study, working a few directories away, reported thirty-six of its own places sitting on a containing place's coordinate and stamped exact — including Berkeley, the town that archive is about, pinned on the geographic centre of England. Its detector is twenty lines: a place has inherited its pin when its coordinate sits on that of another place whose full name is a comma-tail of its own. It was run here rather than assumed not to apply. The atlas came back clean — 165 places, no two on one point, and no precision field to break. The research map came back with four, all in Croatia:

· Corrected in place: /map/

was Published this afternoon on finding «Victorio de Franceschi filio Antonij» in the 1773 marriage register: «A probable identification, left open. This archive carries a Vittorio Defranceschi, 1758–1812, the only Vittorio at the castle, known from his burial. He would have been fifteen in February 1773 — young for a witness … The two are not merged.»

now The burial does not merely give a Vittorio of the right age. It gives his father's name, and the name is Antonio. The row in burials.json reads: Vitorio Defranceschi, born 1758, died and buried 8 November 1812, aged 54, Svetvinčenat, kin: Antonio. It has been in this archive since the graves work went in. So the caution was not caution; it was not having read our own file. Leaving an identification open is the right default when the evidence is a forename and a decade. It is the wrong default when a second independent record names the same father — and the cost of finding that out was one grep, against data already published on this site. A third support turned up in the same row-set. Vittorio's own children are there too, and the son of 1811 is called Antonio — the paternal grandfather's name, in a parish that uses that habit. The two Vittorios are now treated as one man and the note says so. The rule this leaves behind: search the archive before searching the film. An afternoon of frames was spent looking for a baptism to establish a fact that a file in this repository already carried.

· Corrected in place: /who/vittorio-de-franceschi-son-of-antonio/

was «Catharina filia Dominici de Fra[nceschi]», 18 June 1784 — which would have been the first documented daughter of the adult Domenico.

now Magnified: «Catharina filia Dominici de Paz ex Rubino» — a Rovinj family. Three letters where the eye supplied eleven. It was the most valuable candidate of the sweep and it was false, which is precisely why the magnification rule exists.

· Corrected in place: Svetvinčenat baptisms 1764–1791, frame 104

was Three rows placed this family at Svetvinčenat, duplicating rows the archive already carried correctly at Šterna.

now The index entry's image ark resolves to film 005497887, frame 627 — not Svetvinčenat's 005497885. The act reads «Adì 27 9bre 1778. Copich. Felice Ant.o figlio legitimo e naturale di M.ro Bortolo De Franceschi», and the sister record of 1786 is indexed at «Sterna, Litorale Austriaco». Šterna is in north-western Istria, about seventy kilometres from the castle-town. Corrected.

· Corrected in place: Bortolo De Franceschi's household — Felice Anto (1778), Zannetta, Bortolo

was After reading all 127 openings of the volume, this archive was about to state that no baptism of any child of Vittorio and Euphemia appears in it, and that the 1784 index entry must therefore come from some other register.

now The act is on frame 106 — a frame the sweep had already read and passed over. The index entry's own image ark points straight at it. Magnified, it reads «Maria Matthia filia legitima, ac naturalis Victorii, et Euphemiæ jugalium De Franceschi, nata 3 currentis». A complete frame-by-frame read is not a guarantee, and this claim was caught before publication only because the index was checked against it.

· Corrected in place: Svetvinčenat baptisms 1764–1791, frame 106 (4 September 1784)

was «Joanna filia legitima et naturalis Dominici et Euphemiae jugalium Franceschi» — read as an actual baptism of the name, the first in two hundred frames.

now Magnified: «…jugalium BROZINA». Both forenames were right — Dominicus and Euphemia are the two this family uses — which is exactly why it nearly landed. The forenames did the persuading and the eye supplied the surname.

· Corrected in place: Svetvinčenat baptisms 1764–1791, frame 122 (27 July 1788)

was «Dna Helena uxor Dni Dominici Franceschi» — which would have contradicted Fosca and implied a remarriage.

now Magnified: «Dnus Sebastianus Defilipo, et Dna Aglaes uxor Dni Benedicti Fiorenci». Neither name is ours. The reading that would have overturned an established fact was the false one.

· Corrected in place: Svetvinčenat baptisms 1764–1791, frame 120 (22 September 1787)

was That there was «only ever one Albina and one Chinkopan marriage», that it was the sister's, and that the 2012 reunion album had therefore «given Rotilio his sister's husband's surname». Rotilio was left as «the least evidenced of the eleven», and David's tree was treated as carrying a duplicate.

now There were two. The Trieste marriage proclamation of 28 April 1928, Part II Series A no. 132, reads «Defranceschi Rotilo, celibe … nato a Gallignana … figlio di fu Antonio … e di Valhart Maria» and «Cincopan Albina, nubile … nata a Gallignana … figlia di Giuseppe … e di Trompich Maria». Rotilio married an Albina Cincopan of his own village, and his sister Albina separately married Frank Chinkopan. Cincopan and Chinkopan are one surname spelled two ways. The album's detail was correct, David's tree was correct to keep two Albinas, and the inference published here was wrong.

· Corrected in place: /the-eleven/ and /diaspora/ — «So the marriage runs the other way»

was That this archive names living relatives by name only — no date of birth — and that «everyone else on this site is dead and is given in full». The rule was correctly applied to the 51 people David had flagged: not one of them carries a birth year.

now The rule was right and the sentence was wrong. The flag was only ever set by hand, for relatives David knew about — so it caught his family and nobody else. Testing the roster mechanically against the dates rather than against the flag found eight people born between 1931 and 1946 with no death recorded anywhere this archive can reach, every one of them printed with a birth year, and three of them with a full birth date taken from a refugee file: the children of the Dignano household in the IRO «Care and Maintenance» papers of the 1947 esodo, born 1937, 1939 and 1942. They are now held as possibly living — name and place, no dates — under the site's own rule. The dates are still held here and simply not printed, which is what the notice already promised for everyone else. If a death is found for any of them the dates come back. The general failure is worth more than the eight rows. A hand-set flag cannot express «this person may be alive»; only the arithmetic can. The check that found them is three lines — born within living memory, no death, not flagged — and it had never been run. It is now part of the living audit. Said plainly because this is the one rule on this site that is not about accuracy: it is about somebody's privacy, and it had a hole in it for as long as the roster has existed. --- Same day, and the scope was David's to set, not this archive's. Asked whether the rule covers strangers as well as relatives, he answered family only, as written: a person already published by a public index is not made more private by this site withholding what the index prints. So three of the eight were put back within the hour — the Dignano children out of the Arolsen refugee papers, who are not in the family tree. Five remain held, all of them people in David's own tree. The paragraph above described eight; the outcome is five, and the change is recorded here rather than edited into it. The test stands either way, and it is the part that lasts: the arithmetic runs before every audit now, and a hand-set flag is no longer the only thing standing between this site and somebody's date of birth.

· Corrected in place: /people/ — the notice at the head of the roster, and eight rows on it

was That eight households on this page have a Defranceschi woman as the mother — «found by searching the in-law families rather than our own… this is the half of the family the surname loses» — and that the remaining 119 are households of the name.

now Thirty-five of the 127 have a Defranceschi mother and a husband of another name. The eight were the ones built on purpose, by sweeping 33 in-law surnames through the index. Twenty-four more came in through the ordinary baptism sweep and were never re-read — Vellico, Miccoli, Mosscheni, Giachin, Codacovich, Cernobori, Demarin, Delton, Dobrez, Zustovich, Biasiol, Damiani, Zanghirella, Gorlato, Nascovich, Subiotto, Persich, Lazaro, Sandalj, Ruman, Domazetovich, Judović, Baviello — every one of them a daughter marrying out, and every one of them counted among the households the surname continued through. Three more had the wife standing in the father's slot — «Andrina Defranceschi × Guglielmo Defranceschi», «Andrina Defranceschi × Gulielmus Lazzer», «Catharina Defranceschi × Josephus Jelčić» — so a woman was listed as the father of her own children, and the household read as male-line for that reason alone. The classification is no longer a matter of which sweep found the record. A household is a daughter's household when the mother carries the name and the father does not, tested on the data. That moves the split from 119 / 8 to 92 / 35. The sentence «this is the half of the family the surname loses» was the truest thing on the page, and the page was showing less than a quarter of it.

· Corrected in place: /households/ — «The daughters' households», and the line on 27 household records

was That «every row now carries a way out of the table», and that ««What it produced on this site» is the other direction: from a search to the pages its findings ended up on». The paragraph has been on the page since it was written.

now The trail was in the data and on no page. 312 of the 413 rows carry it — 666 links — and the row template passed the field to nothing. The stylesheet even had rules written for it, .sr-ev and .sr-evl, which had never matched an element on the site. It renders now. The rows are drawn by a component in the shared kit, which this archive does not own and did not change; the trail is appended to each row's own HTML instead, which it does own. And nine of the links were dead. /married-in/ is a dynamic route with no index page, /atsea/ was a typo for /at-sea/, and /america/, /listria/ and /families/ are pages that do not exist. They now point at /marriages/, /at-sea/, /diaspora/, /istria-sweep/ and /households/. The site's own link check never saw them, because a link that is never rendered cannot be checked — and it catches them now that they are. Found by asking whether the pages had kept up with the research rather than whether the research was right.

· Corrected in place: /searched/ — «What it produced on this site», and every row that carries one

was «He is one man, and he remarried. The doubt above is resolved.» — that Anzola bore the 1737 twins and died, that Francisca followed from 1743, and that «the two women do not overlap».

now They overlap, and the proof was already in this archive when the claim was written. households.json has carried [h107] Giacomo de Franceschi × Pascasia, four children 1750–1759, read off the same volume at images 155, 163, 169 and 179, since long before this session opened the film. Francisca is Jacobo's wife in 1743 and twice in 1752. Pascasia is bearing his children from 1750. One man cannot have both, so the parish holds at least two Giacomos — and with [h109] Giacomo × Nicolina Teserati (1773–1778) and a «Domenico Gasparo, son of Giacomo» baptised in 1732, it holds forty-six years of the forename fathering children. The original doubt is restored rather than patched. Either more than one Jacobo was living at the castle at once, or the godparent formula lies about husbands' forenames. The first is now the likelier, and the 1738 Agata and Jacoba lines go back to being unexplained. Nothing was misread. Every entry quoted is quoted correctly, and the 1743 baptism is as good a document as this archive owns. What failed was the step after the reading: a new document was laid against the six entries this session had read, and against nothing else. The contradiction was not in an archive in Pazin. It was in a file in this repository, about a volume this project had already been through. The same hour supplied the mirror image, and it is why the practice changes rather than the prose. The twin born beside Antonio in December 1737 had been left deliberately unnamed here — «this archive will not invent a person out of a doubtful word» — while an index row two hundred lines up in the same roster.json read «Casparg, of Jacobo De Franceschi and Anzola, 26 December 1737». One unchecked file cost a wrong conclusion; another cost a right one. Reading a new source now ends with a query against the roster and the households, not with a paragraph.

· Corrected in place: Jacobo de Franceschi of Svetvinčenat, Casparus of 1743, and commit 4100389

was The household list gave a full date of birth and a town for Ursula and for Renate, daughters of Petar and Luise. The dates are not reproduced here, for the reason this correction exists. Beside them, the caption on Ruža's fiche said of her infant son: «he was born in 1944 and may be living».

now Both statements were wrong, and in opposite directions. Ursula and Renate are 86 and 81. This archive's rule is that a living person appears by name and by town and by nothing else — no date of birth — and the list printed both. Their towns, Urach and Metzingen, are kept; the dates are withheld, and both now carry living and bHeldNotPrinted in the roster. The son is not living, and had not been for years. He is Anton Mario De Franceski, born 21 August 1944 at Reutlingen, who sailed to Australia at four and died in 1999. The archive holds his arrival card, his parents' NAA file and his grave — and the caption still said he «may be living». The caution was stale, not careful. This correction was itself wrong for about ten minutes: on finding the caption, this archive redacted Anton's date too and flagged a dead man as living — before reading the row it was editing, which carried d: 1999. Reverted. Why no gate caught any of it. checkliving matches the NAMES of people flagged living against the built pages. Ursula and Renate were never flagged, and the list writes them as bare forenames under a dash — «— Ursula» — with the surname carried by the heading above. A person with no surname on the page and no flag in the data is invisible to a name-phrase gate. The dates were found by grepping for a string while checking something else. And one trap on the way out: the first draft of this notice quoted all three dates in full, to show what had been wrong. A correction that reproduces what it is correcting has not corrected anything — the dates would have moved from one page of this site to another. Never quote a withheld value in the notice that withholds it. One more thing the same file settled. An hour earlier two rows were published from the Arolsen fiches saying whose daughters Ursula and Renate were «is not settled», weighing one camp address against another. pfullingen.json has listed them under Petar and Luise Salzer all along, with their birthplaces. That is the second time today this archive published an open question its own files had already closed.

· Corrected in place: The Pfullingen household list, and the caption on Ruža's camp fiche

was The commit message said the wrong flagging of a dead man as living had been «Reverted», and the script that did it printed Anton restored: b=1944 d=1999 living=None.

now It was reverted in memory and nowhere else. The script loaded roster.json, edited the row, printed the confirmation from the object it had just edited — and had no json.dump for that file. It wrote the two other files it touched and not this one. So the log said the fix had landed, the fix had not landed, and the commit repeated the log. It reached production: for about an hour the deployed site carried Antony De Franceski — who died in 1999, and whose arrival card, parents' NAA file and grave this archive holds — flagged living: true with his birth year stripped. It was found by the landing session, which was rekeying the trees, noticed its own build straddling the correction, and said so. This archive's own verification did not catch it, because the verification was the print statement in the script that failed. A print is not a write. The row is now read back from disk after writing, and the value that is reported is the value that came off the disk. And a gate, because the lesson alone is worth nothing — that was established earlier the same day, when this archive wrote down a lesson about staging and then broke it four commits later. scripts/check-living-data.py now fails the build on living: true together with a death year, and on living: true together with a printed birth year. It was mutated against this exact bug before being trusted: it fails, exit 1, and passes when the row is restored. It also reports every row with a printed birth year, no death year and no living flag — the shape the Ursula and Renate leak had. On its first run it returned four. Three are people found in a public Italian IRO file and correctly keep their dates under the «family only» rule. The fourth was David's father's brother, whose row says «Deceased» in prose and carried no flag a checker could read; he now has one.

· Corrected in place: site/src/data/roster.json — Antony De Franceski, and commit aef6694

was «Two daughters of Petar — born 1940 and 1945 — Urach and Metzingen. Omitted by name: they may be living.» And on the counselling card: «the wife and two children, born 1940 and 1945.»

now This archive corrected this leak four days ago and only corrected half of it. The individual rows had their dates withheld. The summary entry and the prose note kept the years, and nobody looked at them, because the fix had been aimed at the rows. The shape is the part worth keeping. That page deliberately withheld their names and published their years. The roster, equally deliberately, names them and withholds the dates. Each page followed half the rule, in opposite directions — so between the two of them a reader had both halves. A rule obeyed in opposite directions on two pages is not obeyed. Found by a gate written the same afternoon, on its first run. scripts/check-dates-everywhere.py walks all 121 data files rather than the roster alone, because the roster was never where the leak was. It took three passes to stop crying wolf, and each false alarm taught it something true: a death can be recorded in a field (d, deceased), in prose in the same node — Anton carries his birth date in born and «He died on 28 June 1999 and is buried at Warrill Park» in role — or in another file entirely, as Robbie does, with deceased: true in the roster and nothing beside his year in households.json. Miss any of the three and a correct record becomes a false alarm; and a gate that cries wolf on the healthy case stops being read, which is the fault this whole week has been about. Two-digit years are reported and never failed on: «20.1.90» in this register is 1890, not 1990. Negative-tested by putting the exact leak back: it fires, names the entry, exits 1, and passes again on restore.

· Corrected in place: The Pfullingen household list — the entry «Two daughters of Petar», and the note on Peter and Luise's counselling card

was Three rows apiece. For Ursula: «Petar Vid and Luise's elder daughter, born at Urach» and a second Urach row, both read from David's live tree on 15 September; and urzula-defranceski, created from her Arolsen camp fiche on 23 September. For Renate: «Petar Vid and Luise's younger daughter», a second row placed «Metzinger», and renate-defranceski, created the same day.

now Same parents, same birthplaces — Urach and Metzingen — same position in the family. They are almost certainly one woman each, recorded three times. Two of the six are this archive's own work, and they were avoidable. On 23 September it read two fiches, wrote two new people, and never asked whether the tree already held them. The rows that did hold them say «Petar Vid and Luise's elder daughter» and «younger daughter» in as many words. This is the fault this archive spent that same day correcting in three other places — a conclusion published without querying the files already in hand. It was written up as a lesson, a gate was built against a different instance of it, and it was committed again four days later in the one category where it does most harm: living people, each duplicate multiplying their presence on a public site. Found only because a chart caption named renata-defranceski-metzinger while a page was being checked for something else entirely. That is luck, not method, for the second time this week. Nothing has been merged. Who a living person is, is David's to say and not an inference to be drawn from a camp card. A merge performed on a guess is a person deleted on a guess. All six rows now name each other and the question is his. One thing the tree knows and the fiches do not: there is also a son, «Petar Vid and Luise's son, and a person this archive did not know existed». The IRO letter of December 1948 classifies «les deux enfants» — two children — which fits only if he was born after it was written.

· Corrected in place: The roster: Ursula and Renate, daughters of Petar Defranceski and Luise Salzer

was «…needs the Gračišće and Gologorica baptisms of about 1909, which are on film and readable the same way the Svetvinčenat marriages were.»

now Gologorica is not on FamilySearch, and this archive knew. waypoints.json records a walk of FamilySearch's «Croatia, Church Books 1516–1994» (collection 2040054), September 2026 — 1,000 parishes, 24% coverage — and it lists an absent set. Gologorica, Gračišće and Pićan are all in it, along with Vodnjan, Motovun, Lindar and eight more. So the route named in the item does not exist, and it was named without checking a file already in this archive. The Svetvinčenat comparison was doing the work: that parish is filmed, this one is not, and the assumption travelled from one to the other unexamined. What is actually true. The volume that holds a 1909 Gologorica baptism is DAPA no. 1166 — Gologorica, MKR, 1880–1948 — and the inventory says it is held at BA Poreč, the archive of the Poreč-Pula diocese, not at Pazin and not at the parish. It is a letter, not a film. ⚠ The FamilySearch catalogue was queried again today to re-verify from the other side and returned «Something Went Wrong» on every attempt — a service fault, recorded as neither confirmation nor contradiction.

· Corrected in place: Item 160 and Anton Defrančeški's page, both written earlier the same day

was «Antonio's line is worn, and reads 185[5 or 6] to 19[2]4. That is consistent with the chart's 6 August 1856 and 23 October 1924 and does nothing to contradict it…»

now David read the stone itself and gives «ANTONIO DEFRANCESCHI 1855 – 1921». The archive's guess had bracketed the worn digits towards the chart — «185[5 or 6]» and «19[2]4» — which is the reading that agrees with what it already believed. Read on the stone, both ends fall a year or more short of it. And the withdrawal exposes something the archive was already carrying without noticing. This same page quotes the parish Status Animarum: «Antonius Defranceschi q. Antonii, murator, 6-8-1856, m. 23-2-1885, † 23-10-1929». The American chart says 23 October 1924. Both documents give the same birth, 6 August 1856, and the same death day, 23 October — and differ by five years in the year alone, which is the shape of a one-digit copying error. The archive had printed 1924 and 1929 beside each other and remarked on neither. So Antonio now has three death years — 1921 on the stone, 1924 on the chart, 1929 in the register — and this archive asserts none of them. Nothing is merged, nothing overwritten. What settles it is the Gračišće death register, or the obelisk photographed in raking light. ⚠ Maria's death day moves too, from 3 X 1945 to 8 X 1945 — 3 and 8 being the classic pair to confuse on worn stone. Her birth, 6 April 1867, is confirmed three independent ways and is not in doubt.

· Corrected in place: The eleven — Gračišće, the row «What the stone does not settle»

was Both microfilms were held and described as «ZM34K-214 / ZM34K-215 — Pazin Records Office: Gologorica», and the archive's one search-register row for the whole 1,463-frame run reads «the Gallignana baptism book 1667–1745». Cerovlje appears in no file.

now The first 139 frames of reel 214 are not Gologorica. The title card at M01068776 reads «MATIČNE KNJIGE RIMOKATOLIKA · ŽUPA: CEROVLJE · STATUS ANIMARUM · XIX-XX. ST. · SIGNATURA: 3.2». Why it was missed: on the DAPA paper inventory these rows sit above the GOLOGORICA heading with their period column reading only «XIX-XX. ST.» and no parish named on the row itself — so every reading of that page has attached them to whatever heading came next, or skipped them. Cerovlje is the next parish north of Gračišće, on the Pazin–Rijeka road. A status animarum names every soul in every house. Nothing is claimed from it yet — it has not been read — but it is 139 frames of household census this archive did not know it held. (doc — the film's own title card.)

· Corrected in place: The frame map; previously nowhere at all

was «Francisca Defranceschi — Wife of Antonius (house 6). Maiden surname not captured.»

now The 1826 status animarum writes her «Francisca uxor n. Defranceschi» — n. for nata — so she was born a Defranceschi, and this is a marriage inside the name. Recorded as inf, not doc: the abbreviation is read once, and a hand is only legible against itself — it wants a second construction of the same scribe writing n. elsewhere before it is relied on. And the provenance column answers a question this archive had not thought to ask. «Locus ex quo vel ad quem» gives, against the marriage of 29 January 1840: > Antonius — «Ex Parochia Lindaris» · Francisca — «Ex Parochia Gallignana sub N.º 5» Francisca came out of Gračišće house 5 — and this archive already holds that household's census page, DAPA ZM34K-217, M01072071, supplied by the archive on 24 June 2021 and never connected to her. (doc for both places.)

· Corrected in place: The Gologorica line

was Both rows stood on secondary evidence: b. 1768, d. 1829, Gologorica from David's MyHeritage tree with no note at all, and b. 1765, d. 1829, Rovinj from a burial record.

now The Gologorica man is now documented, and his birth is 17 April 1768. The 1826 status animarum enters him as «Petrus Defranceschi», condition civilitas, born 17 April 1768, dying in 1829 aged 61 — a bare year, no day or month. The arithmetic holds: he turned 61 that April. ⚠ The two men are kept apart, deliberately. Different birth years, different parishes, and a shared death year is not evidence of identity. Merging them on the strength of a name and a year is precisely the error this archive keeps catching in its own work. ⚠ And all five Petrus Defranceschi on this roster share one person page, /who/petrus-defranceschi/ — born 1765, 1768, 1772, 1803 and 1889, at Rovinj, Gologorica, Bakar, Vitipolis and Gologorica. The pids are distinct and frozen, so nothing is merged in the data; but a reader arriving there meets five men under one heading. That is a presentation problem, not a data one, and it is written down here so it is not mistaken for the second. Recorded also because the first reading of this entry, at page scale, was 1765 — which would have collided the two rows. At magnification it is unambiguously 1768, and the age at death independently requires it.

· Corrected in place: The roster — petrus-defranceschi-1768 and petrus-defranceschi-1765

was «Two Defranceschi households sit in the 1780s baptism register, both headed by a Joannes Baptista, both with a wife called Mariana, living in different houses — casa 31, carrying the byname «detti Leonardi», and casa 36, recorded as son of the late Josephus.»

now Neither wife was called Mariana. Read in order, the baptism book separates the two houses cleanly by the mother's maiden name: > casa 31 — Joannes Baptista × THERESIA née GRACH (Franciscus Antonius, 13 April 1785; Joannes Baptista, 13 August 1787) > casa 36 — Joannes Baptista × CATHARINA née JABLICH of Buccari (Josephus Antonius, 6 January 1785; Catharina, 5 March 1786; Maria Rosalia, 20 July 1789) That is a better discriminator than a byname, because it is written in the register on every child rather than inferred. The sentence on the line page has been left standing for now and is listed here so it is corrected deliberately rather than quietly. (doc.)

· Corrected in place: The Gologorica line

was Published 1 October: «Where Mathea Defranceschi uxor Leonardi stands godmother in 1805, her Conditio Patrini is written Rust. — rustica. On the same pages the casa 31 and casa 36 Defranceschi are entered Civ. — cives. So the register itself separates the two branches by class.»

now Reading on, the same woman is entered Civis. She stands godmother again in 1810 (M01069066) and 1815 (M01069078), and on both the Conditio column reads Civis, not Rust. So either the condition is being recorded for the other godparent on the row, or it varies with the scribe, or her household's standing changed. The column is a single cell serving two godparents, which makes it ambiguous by construction. The claim is withdrawn to a question. The distinction between a rustic and a civil Defranceschi line may still be right — it rests on the Dominus title and the byname independently — but the conditio column does not prove it, and yesterday's commit said it did. Caught by reading five more folios.*

· Corrected in place: The Gologorica line

was This archive holds Maria Defranceschi, born 16 August 1811, and the 1826 census agrees.

now Her baptism entry says the 14th. «Die 17ma Augusti 1811 · Maria Aloysia Defranceschi nata die 14ta hujus hora 8va mane» — born on the 14th at eight in the morning, baptised on the 17th. ⚠ This is NOT the same thing that explained her brothers. For Carlo and Franciscus Xaverius the archive held the birth and the census recorded the baptism, and both were right. Here the baptism register itself gives a birth date two days earlier than the census gives. Two registers disagreeing about a birth is a different problem, and the primary one was written three days after the event. The date is flagged and not changed, because one reading of one entry does not move a date that two other sources agree on — but the better source now disagrees with them and that has to be visible. Also: her name is Maria Aloysia. This archive carried her as «Maria». (doc.)

· Corrected in place: The Gologorica line

was «✅ All thirty-three skipped frames are now read. The three rows that overstated walked have been corrected, and this range has genuinely been covered.»

now Thirty had been read. M01069049, M01069055 and M01069062 had been listed as outstanding in the previous entry and were then not opened. They have since been read, so the claim is true now — it was not true when it was made. ⚠ This is the second overstatement of coverage in two days. The first marked sixty frames walked when twenty-seven had been opened. Both were caught by this archive auditing itself rather than by anyone raising them, which is the system working; but a fault that recurs within a day is a habit, not an accident, and it is recorded as one. What the habit is: writing the summary before checking the count. The fix is mechanical — scripts/check-walked.py already exists for the walked flag, and the same discipline applies to any sentence beginning «all».

· Corrected in place: The search register, and the commit that announced it

was «Maximilianus Antonius Josephus Defranceschi, b. 9 January 1828, d. 6 October 1910».

now His baptism entry says «natus heri media nocte» — born the previous day at midnight. So the 9th is the baptism and the birth is 8 January 1828, which is the third time this pattern has explained a date in this household: Carlo, Franciscus Xaverius, and now Maximilianus. The archive's dates for this family are baptisms wherever they came from the census. ⚠ And a later hand has written «† 6/8 910» in the margin of that baptism. Read at seven times magnification. In Austrian and Croatian usage 6/8 is 6 August, against the 6 October 1910 this archive holds. The date is NOT changed — one marginal note does not move a date held from another source, but the parish's own margin is well placed to know, and the disagreement is now visible rather than silently resolved. (doc for the birth; the death flagged.)

· Corrected in place: The Gologorica line

was 30 September: a «Franciscus Salman», shoemaker of Gologorica, born 1796, recorded as a coincidence worth knowing — four kilometres from Gračišće and five months from Francisca Salomon's birth. 1 October: downgraded when the 1808 register was found to write Palman, but not withdrawn, because the 1826 reading survived checking.

now The 1825 baptism of his daughter names his father, and that ends it. «Franciscus PALMAN filius Stephani, Sutor», with «Antonia ejusdem uxor nata Brusich» — the same couple the 1826 census enters as «Franciscus Salman, Sutor» and «Antonia uxor n. Brusich». A patronymic is not a spelling. It ties him to the Stephanus Palman who runs through this register from 1799 — children Theresia (1799) and Lucia (1808), wives Sancta and Catharina née Jurich. Three entries write P and give a father; one writes S. So there is no Salman at Gologorica, and the resemblance to Salomon was an artefact of a single scribe's capital. The lead is withdrawn in full, on the page where it was published. Published, downgraded, withdrawn — three days, three states, all three visible. That is what the grade is for. (doc.)

· Corrected in place: The Salomon maternal line

was Published earlier today, after reading to folio 97: «At the end of 1833 the register changes tongue. Folio 96 opens «1833 adì 23 Xbre» and never looks back.» The section was titled for a year, as though 1833 were a boundary in the book.

now It looks back on 6 May 1835. (doc — DAPA ZM34K-214, M01069111, folio 98.) In the middle of that page the entries return to Latin, and the hand that writes them signs itself «Petrus Milcetich, Parochus» — one of the eight priests the 1826 status animarum lists, born at Omišalj on Krk. The Italian spell lasted seventeen months and it belonged to one man: Marco Marussich, Amministratore Parrocchiale, who held the book from December 1833 to April 1835. The corrected finding is stronger than the one it replaces. The language of this register is not a date. It is a signature. A reader who learns «Italian after 1833» will miss every entry from May 1835 on; a reader who knows the language tracks the hand will look for both forms at every point and be right. ⚠ And the error is the same shape as this morning's other one: a claim generalised from the frames in hand rather than restricted to the frames it covers. «never looks back» was an assertion about 44 unread folios made on the strength of two. The rule it adds: a sentence about what a book does FROM a date is a claim about every page after it, and may not be written until those pages are read — or must say how far the reading goes.

· Corrected in place: Four more hands, and the year the register changes language

was Carried from the family tree since this roster was built: «Joanna Defranceschi, born 9 June 1848, Gologorica», daughter of Antonius Defranceschi and Francisca — with no death year.

now «Junii 8ᵃ · [domus] 5 · ROSA nata heri», of «Antonius Defranceschi» and «Francisca nata Defranceschi», godparents «Giovanni L— et Adelina Defranceschi», entered cives. Born 7 June 1848, baptised the 8th. (doc — Gologorica baptisms 1784–1880, DAPA microfilm ZM34K-214, M01069123, folio 109.) The name is wrong and the date is two days out, and it is one child rather than two. A mother cannot bear children on the 7th and the 9th of one June; no second entry stands anywhere near it; the parents, the house and the week are the same. The register is the document and the tree is not. ⭐ And the same line closes her life, which the archive did not have at all. A later hand has written «† 10/12 99» beneath the entry — she died 10 December 1899, aged fifty-one. The pid is kept. joanna-defranceschi-1848 will go on reading «joanna» for ever, because a moved id is a dead URL and the address was published. The row carries her real name; only its id remembers the error. Her duplicate row, joanna-defranceschi-1848-2, is renamed with her. ⚠ The day was read with the rule learnt this morning — crop past the cell, not to it. The «8ᵃ» of her baptism was set against the «9ᵃ» of the row below it, where the 8 is a double loop and the 9 a single loop with a descender.

· Corrected in place: Her page · the Gologorica line · the pedigree

was Published earlier today, on five visitation certificates reading «Vidi. PETINAE»: «Fifty years after the diocese of Pićan was suppressed, its deanery still sits there and the dean signs there» — and the worklist was given it as «an address, not a guess».

now A sixth certificate, of 15 March 1849, reads «Vidi. PIZINAE» — PAZIN. (doc — DAPA ZM34K-214, M01069123, folio 109.) Cut from the film and set beside the 1838 one: the 1849 word has a dotted i and a z; the 1838 word has an undotted e and a crossed t. They are two different towns and both are in this book. So the dean signed at Pićan in 1838–1844 and at Pazin in 1849. *Whether the deanery moved, or the dean simply signed wherever he happened to be sitting, this archive does not know — and that is the point: one place-name repeated five times looked like an institution, and the sixth shows it was a habit. What stands: the visitations themselves, their dates, and the named deans — Franovich (1804), Anicich (1808), Coverlizza (1838–1859). The series is real and annual. What is withdrawn: the seat, and with it the instruction to direct an enquiry to Pićan rather than Pazin. Both towns are now named on worklist 169 and neither is preferred. ⚠ The lesson is the same one this archive learnt about «never looks back» eight hours ago, in a different costume: a claim generalised from the instances in hand rather than restricted to them. Five readings of a word are not the word's definition.

· Corrected in place: Worklist 169 · the Gologorica line · the search log

was David asked on 5 October whether everything found that day had reached the site — «every tree, person, line, story». The honest answer was no, and it took an audit to find out rather than a memory.

now Three gaps, all of the same shape: a finding written into the prose and never carried into the data the pedigree is drawn from. 1. Maximilianus Antonius, born 8 January 1828. The archive established on 4 October that the 9th is his baptism — «natus heri media nocte» — and wrote it on his own page in those words. The pedigree went on showing 9 January as his birth for a day. Corrected. 2. Adelaidis was placed and never added. On 5 October «Maximilianus Defranceschi ET SOROR ADELAIDIS» (1847) and «Enrico Defranceschi ET SOROR Adelaidis» (1852) moved her out of «unplaced» and into Josephus and Lambertina's household — on her roster page, in the story, and in two commit messages. She was not in the family. Added, flagged with the open question of whether she is the same child as Anastasia or Amelia. 3. The elder Clementina of 1821 was never in the family either. Her baptism has been cited here since 1 October and her story — the reused name, and whether she died a child — is told in three places. The pedigree showed her parents with one Clementina. Added. ⚠ WHY THIS CLASS OF ERROR IS DANGEROUS. None of it is a wrong fact. Every one of these was published correctly somewhere a reader could find it — so no link broke, no check failed, and check-propagation is silent, because it looks for a WRONG claim left standing and this is a RIGHT claim that never travelled. The pedigree is the one thing on this site a reader trusts without reading the prose, and it was quietly a day behind the prose. What was done: a direct audit — every person documented that day checked for a built page, a citation and a place in the family; every tree record checked for agreement between its written date and its year field (0 disagreements); every story page checked for the findings that belong to it. It also found that the Peschle findings and the surviving Salamon lead existed only on the Gologorica page and the worklist, and never reached the family pages they belong to. Both now carry them. The rule this adds to the method: a finding is not propagated when it is written in prose — it is propagated when it is in the data the diagrams are drawn from. Ask of every finding: which generated view would be wrong if I stopped here?

· Corrected in place: The Gologorica pedigree · Maximilianus · Adelaidis · Clementina 1821

was Carried on the pedigree and in the roster: «Hermenegildus Defranceschi, 1865, Gologorica» — a son of Franciscus Xaverius Defranceschi and Clotilda Ortis.

now «1865 Aprilis 23ᵗⁱᵃ · [domus] 7 · BEATRIX HERMENGILDIS n. 10 Apr.», of «Franciscus Defranceschi, Possidens» and «Clotilde n. Ortis»; godparents «Luigi Ortis» and «Ermina Marina», entered cives; midwife «Joanna Crismanich». The Sexus column marks her female. (doc — DAPA ZM34K-214, M01069142, folio 128.) Born 10 April 1865, baptised on the 23rd. The date this archive held was right. The name and the sex were not. ✔ And the correction joins two rows that were each half right. This archive also carried beatrix-hermenegildis-de-franceschi-1865, imported from David's live MyHeritage tree with the right name, no place, and the line «unplaced». The pedigree row had her household and her date and the wrong name and sex. Between them the register settles both: the two are marked sameAs, neither id is moved, and the MyHeritage row moves out of «unplaced» into the carnia line. ⚠ The pid keeps the masculine form. hermenegildus-defranceschi-1865 will read so for ever, because a moved id is a dead URL and the address was published. Only the id remembers the error; data/decisions.json carries the re-declaration and the reason. The lesson is small and worth saying. A pedigree that shows sex is making a claim about a person on every row, silently, in a place no prose covers. This one had been making the wrong claim about a girl since the tree was imported, and nothing in the archive could have caught it but her own baptism entry.

· Corrected in place: Her page · the other row the tree carried · the pedigree

was Carried on the pedigree and in the roster: «Robertus Franciscus Josephus Defranceschi, born 21 November 1860, Gologorica», with no death.

now «1869 Decemb. 5ta · [domus] 7 · ROBERTUS FRANCISCUS JOSEPHUS die 21 Nov. natus», of «Franc. Defranceschi, possidens» and «Clotildis nata Ortis»; godparents «Albertus Peschle» and «Maria Defranceschi», cives; and beneath the entry, in a later hand, «† 6/8 1934». (doc — DAPA ZM34K-214, M01069148, folio 137.) Born 21 November 1869, baptised 5 December, died 6 August 1934, aged sixty-four. ⭐ What makes this correction safe rather than a judgement: the pedigree's day and month are exactly right — 21 November — and only one digit of the year is wrong. A wrong year with a right day is a copying slip; a wrong day as well would have raised the question of whether it was the same child at all. And it untangles two children the archive had both standing in 1860. Anna Maria Francisca is the 1860 birth — 28 October, confirmed by the same register in a September correction — and Robertus is 1869. The household had one child in 1860 and one in 1869, not two in 1860. The pid keeps the year it was minted with: robertus-franciscus-josephus-defranceschi-1860 will read so for ever, because a moved id is a dead URL. data/decisions.json carries the re-declaration.

· Corrected in place: His page · the pedigree

was This roster gave Clementina Defranceschi, born 28 January 1821, a death of 1870, carried over from the family tree — while recording as an inference that she had probably died in infancy, because her parents gave the name again twenty months later.

now «1821. Die 30ma Januarii hora ½ 4ta a meridie · [domus] 104 · CLEMENTINA FILIA D.ni JOSEPHI DEFRANCESCHI · dierum 2 et horarum 11½ · Imperfecta Structura», buried 1 February 1821. (doc — DAPA ZM34K-214, M01069259, page 62 of the Gologorica death register.) Her baptism has her born on 28 January 1821 at four in the morning, at the same house 104. From then to half past three on the afternoon of the 30th is two days and eleven and a half hours. The arithmetic closes exactly. ⭐ The inference was right and is now a document — and so was the caution published with it: her baptism carries no marginal cross, and a blank margin in that book is not evidence of a life. ⚠ The 1870 is removed, not moved. It cannot be hers. It does not automatically become her sister's either: the thirty-year dispute over the Clementina of 1822 — archive 1840 against tree 1870 — stands, and has merely lost the comfortable explanation that the two dates belonged to two women. One of them is still wrong about a real person, and the death register has not yet reached her. ✔ One thing it does settle. The Clementina Defranceschi who stood godmother at Gologorica on 20 February 1840, entered civis, was published as possibly either sister. She can only be the younger.

· Corrected in place: Her page · her sister's · the pedigree

was The roster carried «Aloysius Defranceschi, 1832–1870», and on 5 October, when Gisella Elvira was found in the baptism register with a father written «ALOYSIUS DEFRANCESCHI, MERCATOR», the archive recorded the collision and refused to resolve it: «this archive carries Aloysius Defranceschi (1832–1870), Carlo's brother, as dead three years before she was born … nothing is changed on his record». His birth was carried as 8 May 1832.

now Two pages of the same reel, read 11 October, put him alive and at work long after 1870. > Gologorica Status Animarum 1921, house 87 (DAPA ZM34K-214, M01069915): «De Franceschi Aloisius» · conditio «mercator» · n. «4 Maji 1832» · conf. 15 Octobr. 1841 — with «Joanna ux. n. Udovičić», n. 19 Novemb. 1845, and «Gisella, filia», n. 15 Aprilis 1873 > «Senes · Annus 1909» (M01069823), the parish's aged at 31 December 1908: «Defranceschi Aloisius, house 87, born 4 May 1832, aged 76» The merchant, the wife, the trade and the daughter's day all match the baptism entry that raised the problem. The death year 1870 is withdrawn — a man counted among his parish's living at seventy-six in 1908 did not die in 1870. ✔ This archive's other row for him, the MyHeritage «Aloysius / Aloisius 'Luigi' Defranceschi, 1832–1912, Volosca», is consistent with the 1908 list; the two rows are one man, and both ids are kept and frozen. ⭐ And the birth day is changed, on the third reading. The 1826 status animarum gave «4 Maji 1832» against the roster's 8 May, and the archive declined to move it on one page-scale reading. There are now three register statements of 4 May 1832 — the 1826 book, the 1908 «Senes» list and this census, the last two read at magnification — against one tree-derived 8 May. The roster now says 4 May. ✔ And Gisella's father is no longer a dead man. (doc — DAPA ZM34K-214, M01069915 and M01069823.) ⚠ And it is not two men. Worklist 177 left both doors open — «either there were two Aloysius Defranceschi at Gologorica, or the 1870 death belongs to somebody else». The 1826 status animarum shuts the first. It enters this Aloysius in house 7, the household of Josephus Defranceschi and Lambertina née Peschle — so he is Carlo's brother — and gives the same 4 May 1832 as the 1908 list and the 1921 census. One man, one birth day, attested three times in one parish. ⚠ The 1870 is withdrawn, not reassigned. It came off the family tree and no burial entry has ever been found for it; this archive's own standing warning on these collisions is that a tree date is second-hand to the death itself. ⚠ Corrected within the day: the death register is NOT unwalked — this archive read it end to end on 6 October (M01069226–M01069297). That walk read every entry and turned up no Aloysius Defranceschi dead on 25 November 1870. ⚠ The claim rests on the walk, not on the book's index: the index stops at page 71, about 1830, and is silent about 1870 for the same reason it is silent about the two deaths this archive corrected that day. But it ends 23 January 1880 with «Segue libro nuovo · Vol. IV», so it could not hold a death after that date either way; its silence rules out an 1870 burial at Gologorica and nothing more. This row therefore carries no death year at all rather than a wrong one, and Volume IV — not on these reels — is where a later death would be.

· Corrected in place: Aloysius Defranceschi · Gisella Elvira · The Gologorica line

was On 1 October 2026 this archive read the godmother of Clementina Maria Defranceschi, baptised 6 October 1822, as «Amalia PESCHLE ex SILESIA» (DAPA ZM34K-214, M01069089). An entire line of work was raised on that one word. Worklist 170 — «SILESIA — and the possibility that Carlo's mother's family is German» — asked whether the Peschle were German-Silesian who arrived on the Kvarner rather than a Kvarner family; it named Acta Publica and the Zemský archiv v Opavě as the instruments, and directed that any search run Peschel, Peschke and Pöschl as well. The claim spread: to the Peschle page («AND THE FAMILY REACHES SILESIA»), to the plate caption, to four search-log rows and their tags, and into the reading of two later finds — the German-named godparents «Schrott» and «Walser», and the proxy godfather «Anselmus Peschle» — each of which was weighed as possible corroboration of a Silesian origin.

now Re-read at four times magnification on 11 October 2026, the word is «ex VOLOSCA». > V — o — l — o — s — c — a. Seven letters, opening on a flat-topped V and closing on a long terminal a. There is no dotted i anywhere in it and no e. «Silesia» is S-i-l-e-s-i-a and would need two dotted i and an e; the word shows neither. It is not a near thing and it does not need a comparison hand to settle — but the same book supplies one anyway: «Eugenia Peschle VOLOSCENSIS, Dni Mathiæ filia», godmother at M01069058, and Lambertina née Peschle herself entered «Ex Volosca» in the 1826 census. ⚠⚠ WHAT THIS COSTS AND WHAT IT SAVES. Worklist 170 is withdrawn, not narrowed. There is no «ex Silesia» entry in this archive's evidence, so there is no Silesian question — the premise was a misreading and nothing else ever supported it. The planned sweep of the Austrian-Silesian registers is cancelled before it was run, and the instruction to carry Schrott, Walser and Anselmus «into any Acta Publica sweep» is withdrawn with it. ✔ And the correction returns the family to where this archive had it all along. The Peschle page files them as a Kvarner family — «Volosca, on the Kvarner, and Fiume». The register agrees. The «Silesia» reading was the anomaly, and for ten days it was treated as the discovery. ⚠ The lesson is the one already in the rules, and it was not followed here. A doubtful word is settled against the same hand writing both candidates, at one magnification — not by reading it once at page scale and building on it. This word was never magnified until today. The plate is here. (doc — DAPA ZM34K-214, M01069089.)

· Corrected in place: The Gologorica line · The Peschle · Worklist 170 · The search log

was Earlier on 11 October this archive withdrew a death year of «25 November 1870» from Aloysius Defranceschi, Carlo's brother, because the parish counted him alive at seventy-six in 1908 and heads house 87 with him in 1921. That correction said the date could not be explained: «The 1870 is withdrawn, not reassigned … it came off the family tree and no burial entry has ever been found for it.»

now It is a marriage date, and the register gives it in the marriage column. > Gologorica Status Animarum 1862 (M01069542): «Aloysius Defranceschi», mercator, n. 4 Maji 1832; «Joanna n. Udovičić», n. 19 Nov. 1845 — and in the Matrimonii contracti column, braced across the pair, «25. Nov. 1870» The family tree took the day he married Joanna Udovičić and recorded it as the day he died. That is why no burial was ever found on it: there is no death on that date, there is a wedding. And it is why the collision with his daughters born 1873 and 1878 looked so stark — the archive was reading the beginning of the family as the end of the man. ✔ The marriage is new to this archive, which held no wedding date for them; their eldest, Armella, follows on 7 April 1871. (doc — DAPA ZM34K-214, Gologorica Status Animarum 1862.)

· Corrected in place: Aloysius Defranceschi · Joanna Udovičić · The Gologorica line

was On 11 October this archive published «Marianna ux. n. Grah-Jopan» from the Gologorica Status Animarum of 1862 (M01069464), and called «Jopan» a second compound of the name, new to this archive — reasoning from it that «the Grah of this corner of Istria distinguished themselves by at least two second elements» and that «a search for Grah-Mišan alone would miss her entirely». It went onto the Grah record and into the worklist. ⚠ It was read off a four-frame tile at page scale and never magnified.

now Re-read at nine times magnification, against this scribe's own capital J. The word «Joannes» appears twice on the same page. Its J loops wide at the top left and descends below the line with a leftward tail. The letter after the hyphen in the disputed word has a straighter, taller ascender and does not match it. And the word runs off into the column rule, so its ending cannot be read at all. > What the page securely gives is «Grah-» and a second element beginning with a tall ascender. «Jopan», «Šopan», «Žopan» and «Sopan» are all open. ⭐⭐ AND THE POSSIBILITY THIS OPENS IS WORTH MORE THAN THE CLAIM IT CLOSES. This archive already holds a Grah who was a village headman: «Mattia uxor Suppani Joannis Grach» at Gračišće (M01070111) — and that word was itself recovered from a misreading. «Suppanus» is the župan, and an office is exactly what this region turns into a household byname, beside «Rabar-Baron» and «Zec-Kočebar». ⚠ If the second element reads «Žopan», the compound is not a second family but Grah the headman's house. Nothing establishes that; it is recorded because the letter is ambiguous and the office is attested for this surname one parish away. ✔ What is unchanged: Marianna exists, is a Grah by birth, married Joannes Bubić on 22 January 1861, was born 19 September 1834 and died 19 August 1870. Only the spelling of her father's house is in doubt. (doc — DAPA ZM34K-214, M01069464 and M01068790.)

· Corrected in place: The two Grah wives · The Grah record · Worklist 218

was Twice on 11 October this archive got this wrong, in opposite directions. First it published «Marianna ux. n. Grah-Jopan» as a new compound. Then it withdrew the reading — rightly, the initial does not match the scribe's J — and raised, as an explicitly unestablished guess, that the element might be «župan», an office. It also said the question could only be settled by the same compound in another hand, and that Marianna «bore no child this archive has found».

now The baptism register settles all of it on one page. > Page 126 (M01069140), Patrini, 1864: > 7 Maji — «Joannes Bubich» and «Marianna n. GRAH-MISSAN» > 22 Septembris — «Josephus Udovichich» and «Rosa GRAH-ŽIPAN» Three months apart, in one hand. ✔ The second compound is real and it is «ŽIPAN» — at eight times magnification a tall looped capital followed by i-p-a-n, the Croatian župan, a village headman. The guess is now a reading, and it joins the «Suppanus Joannes Grach» this archive already holds at Gračišće. ⚠ The form is certain; identifying this household with that Gračišće one is not. ⚠⚠ AND MARIANNA IS «GRAH-MISSAN». THAT PART WAS MY ERROR. The register gives her twice — as a mother (M01069137: «Joannes Bubich, agr.» · «Mariana Grah-Missan») and as a godmother (M01069140, with the same husband). The 1862 census reading of her as a «-pan» form is withdrawn: that cell was cramped, was never resolved, and the baptism register says otherwise plainly and twice. ✔ So the Grah women of Gologorica are three, not two — Marianna and Rosalia, both Grah-Missan, and Rosa Grah-Žipan, new. ➕ And «she bore no child this archive has found» is withdrawn too: she bore one, and that baptism is the only reason her maiden surname survives — her burial entry gave her husband's. (doc — DAPA ZM34K-214, M01069137 and M01069140.)

· Corrected in place: The two Grah wives · The compound · The Grah record

An hour ago this archive published that the Gologorica death register's 1773 holds eight deaths and «not one of them a Defranceschi», and concluded that the Leonardo who died in 1751 was therefore the stronger candidate for the 1774 groom's father.

⚠ The eight were eight WOMEN. This volume is split by sex and this archive had already seen it and not drawn the inference. The baptisms in the same book are headed «Notatio Baptizatorum Masculorum a 1ᵐᵒ Julij 1770» and «Notatio Baptizatorum Fœminarum» — a split this page recorded yesterday — and the deaths follow the same division: men at M01068975–987, women at M01068988–996. The frames read were the women's.

⭐ Leonardus is in the male sequence, forty frames earlier:

> «45 · Die 14 Martij obijt · Leonardus Defranceschi annorum septuaginta sex, provisus Sacramento Pœnitentiæ per me Marcum Bancich Parochum, sepultus vero est 15 in Arca D.ⁿⁱ Petri Defranceschi in hoc Parrochiali Ecclesia existente. ( : professionis suæ TEXTOR EX CARNIA : )»

(doc — DAPA ZM34K-214, M01068977, «Liber Mortuorum II», male sequence, 14 March 1773.)*

So both the nil and the conclusion drawn from it are withdrawn. The 1774 groom's father is the man of 1773 — right house, right age, right trade, fifteen months before the wedding — and that identification moves from inf to doc.

⚠ The lesson is the one the archive keeps relearning and did not apply here: when a book's baptisms are split by sex, assume its deaths are too, and look before counting. A nil taken from half a register is not a nil.

· Corrected in place: /gologorica-line/ · /person/leonardus-defranceschi-1697/ · /searched/

Yesterday this archive gave Margarita Defranceschi her death from the obit calendar — «20. · Margarita figlia q. Sig.ʳ Giuseppe Defranceschi, N.º 36 — 1783» — and computed her age from the family tree's 1747, calling her «thirty-five» at a witnessing in November 1782.

⚠ The contemporary register has now been opened.

> «Die 22. Aprilis anno 1783. Margaretta Ancilla annorum 22, mensium 9, filia Josephi et Hellenæ jugalium de Franceschi, provisa Sacramentis omnibus in sua infirmitate a me Parocho; obdormivit in Domino, cujus corpus die 25 ejusdem mensis sepultum est in Arca Parentuum per R.ᵐᵘˢ Archipresbiterum et Parochum CHERBUNÆ D.ⁿᵘˢ Joanes Gràch.»

(doc — DAPA ZM34K-214, M01068991, «Liber Mortuorum II», under 22 April 1783 — the contemporary register, read 10 October 2026.)

Four things change, and a fifth is settled.

⚠⚠ AND IT IS A MEASURED ERROR IN THE OBIT CALENDAR. Yesterday this archive published that the calendar passed three controls, then corrected that to «it is a copy, so the controls prove less than claimed». Here is a case where the copy is simply wrong: two days out on a death it transcribed. That is not a reason to stop using it — it is the only route to the registers it digests — but every date taken from it now carries a known tolerance, and where the register survives the register wins.

⚠ And one thing the register does NOT say. It calls her «filia Josephi et Hellenæ jugalium» — no «quondam». The calendar, written in 1800, calls her «figlia q. Sig.ʳ Giuseppe». The tree has Josephus dying in 1782 — which a scribe in April 1783 would normally have marked. Not resolved here; recorded because it is the second figure on that man's row to come from a tree and nothing else.*

⭐ What survives untouched: [her identification as the «Margare[t]ha de Franceschi» who witnessed on 9 November 1782](/person/margarita-defranceschi-1747/) — an unmarried daughter of twenty-two standing at a neighbour's wedding is, if anything, likelier than one of thirty-five.*

· Corrected in place: /person/margarita-defranceschi-1747/ · /gologorica-line/ · /searched/

The frame map of 30 September was built to replace the Pazin paper inventory, which it showed to be wrong twice, and it stated its own rule:

> «The films carry their own printed title cards, one before each book, naming the župa, the register and the years. Those are the authority, and they are what this table is built from.»

⚠ The card at M01069199 reads «ŽUPA: GOLOGORICA · MATIČNA KNJIGA UMRLIH 1646-1806 · SIGNATURA: 5.1». The book it introduces is not a death register.

Its own title page, three frames later, says it is a «Catalogus mortuorum … ADULTORUM … coactus a me Antonio Ferdinando M[?]ocovrich, Parocho Loci, Anno Domini 1800» — a perpetual obit calendar, compiled in 1800 out of the registers. The item has now been inventoried frame by frame and all twenty-five are accounted for: title card, test target, cover, title page, twenty openings of calendar, endpapers, KRAJ. No narrative register is in it.

What is corrected: the rule. A printed card made by a filming project in 1997 is a third-hand description of a book, and here it is wrong in exactly the way the paper inventory was. The book's own title page is the authority. The frame map's row is amended to say what the item actually is; the frame ranges in that table are unaffected and were never in doubt.

⚠⚠ And the consequence is larger than the correction. This archive has believed since 30 September that it holds Gologorica's deaths from 1646. It holds an 1800 digest of them. The pre-1784 death registers of this parish are not on these reels at all — so the eleven Defranceschi entries the calendar gave up cannot be expanded into fuller entries, because there is none here to go to. That raises what the calendar is worth and closes the work-list item that asked where the register sits.

· Corrected in place: /gologorica-line/ · /searched/

This archive swept the Gologorica obit calendar today and published, of three entries it could check against the parish death register:

> «Three controls, three passes. A finding aid compiled after the fact must be tested on cases you can check before its uncheckable ones are believed.»

⚠ Then the title page was read, three openings further back.

> «Catalogus mortuorum in Parochia Gollogoritiæ ADULTORUM, ab anno Christianæ 1646 usque ad nostra tempora, ex antiquis et recentioribus paginis COACTUS A ME ANTONIO FERDINANDO M[?]OCOVRICH, PAROCHO LOCI, Anno Domini 1800.»

— «Catalogue of the dead adults of the parish of Gologorica, from the year 1646 down to our own times, gathered by me, Antonio Ferdinando M[?]ocovrich, parish priest of this place, out of the old and the more recent pages, in the year 1800.» ⚠ The compiler's surname is read «Mocovrich» and the initial letter is not certain. (doc — DAPA ZM34K-214, M01069202, the title page.)

The calendar was compiled in 1800, out of the registers, by the parish priest. So for every entry before 1800 it is a COPY of the book it was being tested against — not an independent witness to anything.

What the controls actually show: that Antonio Ferdinando copied accurately. That is worth knowing and it is not nothing — it is why the nine entries that cannot be checked can be used at all, since they come from registers this archive has not got. But it is not what was claimed. Three sources agreeing is evidence; one source and two copies of it is one source.

⚠ And it weakens the priest's year. [This archive wrote that the calendar's 1797 is «explicit where the register's [1798] was editorial»](/person/joannes-mathias-franceschi-gologorica/). Both are now readings of the same underlying register — one by a modern archive off a page heading, one by his successor in 1800 off the entry. Mocovrich had the book and probably the memory; he was also copying a century and a half of entries by hand. The tension stands and neither figure moves, but the reason to prefer 1797 is smaller than it was.*

⭐⭐ AND THE TITLE PAGE CARRIES A SECOND THING THAT MATTERS MORE THAN THE CORRECTION: «ADULTORUM». The catalogue is of adults only. So every nil drawn from it is a nil for adults — the Grah nil across sixteen openings says nothing about Grah children, and a family that buried only infants here would be invisible in it. That qualification has been added wherever this archive leaned on the book.

· Corrected in place: /gologorica-line/ · /searched/ · /person/sancta-defranceschi-d1792/ · /person/antonius-defranceschi-1742/ · /person/helena-defranceschi-1725/

The Bakar household page carries this sentence, written when the household was reconstructed from the FamilySearch indexes:

> «A separate entry has a «Joannem Defranceschi» marrying «Catharinam Sablich» — either the same man remarried, or a brother.»

⚠ The register was opened today and he is neither.

> «Die 10 9bris pro domo D. Sebastiani Sablich, Nuptiis in facie Ecclesiæ contractis inter D.ⁿᵘᵐ Joannem Bap.[tistam] [De]franceschi ex GOLLOGORIZZA et D. Catharinam Sablich, tradidit D. Parochus Buccarensis Jacobus Polian, præsentibus Senatore D. Josepho Walis et D. Josepho Smich»

(doc — FamilySearch, «Croatia, Delnice Deanery Catholic Church Books, 1571–1926»: Bakar, Religious Marriage Registers 1739–1808, film 005494714, item 1, image 172 of 226, page 161. Index 1:1:6Z2F-LSSQ; image 3:1:3QSQ-G99X-RX3F.)*

He is not a Bakar Defranceschi. The scribe wrote his village into the entry: «ex Gollogorizza». He is the Joannes Baptista of Gologorica house 36, and the bride is Catharina, whose marriage this archive has been looking for since 30 September.

⚠ How the error was made, which is the useful part. The entry was read from the index, and the index records only «Joannem Defranceschi» and «Catharinam Sablich» — no place. With no place in front of it, a Defranceschi marrying at Bakar in a set of Bakar Defranceschi records can only look like a Bakar Defranceschi. An index strips the field that would have told you the record is not about the family you are reading it for.

What the Bakar household keeps: its own Akačić marriage, its Franciscus, its children of 1773 and 1776, and its Catharina born 1734. What it loses: a second marriage it never had. The age arithmetic that first separated the two Catharinas was pointing at this before the film was opened.

⚠ A second thing this unsettles and does not settle: her surname. Gologorica writes her «nata JABLICH ex Buccari»; Bakar writes «SABLICH» twice, for her and for her father. Neither is declared correct here. The test is the usual one — the Bakar scribe's own unambiguous S and J on the same page, at one magnification. The person id stays frozen either way.

· Corrected in place: /crikvenica-house/ · /gologorica-line/ · /person/catharina-defranceschi-jablich-gologorica/

Published this afternoon, corrected the same afternoon, by reading two more frames of the book it came from.

The claim was that the Defranceschi priest held Gologorica «continuously from 12 November 1782 to at least 22 November 1785». It rested on three entries in which he signs «per me» — and on nothing in between, because nothing in between had been read.

⚠ M01069004 has, between two of them:

> 1 February 1785 — «Matheus Crosgniak viduus ... copulatus est cum Lucia Covacich ex Cerouglie ... per Parochum Josephum Deprato in facie Ecclesiæ»

Three weeks after the Franceschi entry of 12 January 1785, and nine months before the one of 22 November, a DEPRATO is the officiating «Parochum». (doc — DAPA ZM34K-214, M01069004.)

What survives: the three Franceschi entries are unchanged, and he is still demonstrably parish priest on those dates. What is withdrawn: the word continuously, and with it the tidy bracket it was used to build — «so Gallignana falls before 1782 or after 1785». That inference is withdrawn too, because a cure this archive now knows two men were styled to within three weeks of each other cannot be used to exclude a third parish from the gaps.

⭐ And this archive had written the warning itself. The standing rule for Gologorica is that a «De—» surname here may be Defranceschi, Deprato or Defranich, and that the death book shows «DEFRANCES[CHI], Presbyter Loci» and «D. JOSEPH: DEPRATO, ex Parochus» alternating in one column. The marriage register shows the same two men alternating in the same office — and the claim went out anyway, built on the Defranceschi entries alone because those were the ones that had been looked for. A method row has gone up: read the officiant column for everybody, not for your own surname.

The reading of the three Franceschi entries was itself checked at magnification and stands; what failed was the inference drawn across the gaps between them, which is a different kind of error and the more common one here.

· Corrected in place: /gologorica-line/ · /person/joannes-mathias-franceschi-gologorica/ · /searched/

Worklist 191 concluded, on 6 October, that of the two figures in the census column «Nrus. Domus» the upper one was live. The argument was clean: «taken down the book the upper figures climb — 28, 29, 30, 31, 32, 33 … — while the lower ones are scattered», and «a census is bound in the order of its own numbering, so the ordering figure was live in 1826».

⚠ Read at twenty times magnification, the upper figure has a diagonal stroke through it. Three cells, plated: 20 struck above 27 (p.19, Domus Ecclesiæ); 35 struck above 64 (p.43); 30 struck above 64 (p.44). The struck figure is the superseded one. doc.

⚠ And the premise fails too, which matters more than the conclusion. This book is bound in the order of neither numbering. Page by page the uppers run 20, 29, 12, 22 … 35, 30 and the lowers 27, 30, 13, 14 … 64, 64. Neither climbs. An argument from an ordering that is not there cannot support anything — and it happened to be reversible, so it produced a confident answer that was exactly backwards.

⭐⭐ WHAT THE CORRECTION OPENS IS WORTH MORE THAN WHAT IT CLOSES. Pages 43 and 44 carry the same lower number, 64, with «Nrus. Inquilinorum juxta eorum familias» 1 and 2, and different struck uppers — 35 and 30.

> Two separately numbered old houses became two tenant families of one new house.

So the renumbering is many-to-one, and the three days spent looking for a constant difference were looking for something that does not exist. It also dissolves the contradiction this archive had logged as open: «36 vetus · 14 novus» running one way while the census pairs ran the other. A mapping that merges houses is not monotonic, so both directions occur and neither is an error.

⚠ This session's own first readings of those two cells were also wrong — «28» and «36» at the magnification that fits four pages on a sheet, against 35 and 30 at twenty times. They were withdrawn before publication rather than after, which is the only reason they are a footnote here instead of a fourth correction.

· Corrected in place: /worklist/ · /gologorica-line/

Published an hour ago, closing the Gračišće parents-column nil: that the family was «a household of adults, present and in demand, baptising their children somewhere else — and the somewhere else this archive already holds is Gologorica casa 31, five kilometres away, baptising in exactly these years».

⚠ It is not, and the archive had already established that it is not. The Gologorica baptism book was read folio by folio on 1 October and the row says, in its own words:

> «No Defranceschi child is baptised between 1788 and 1799 in this book — the two houses had their children in 1785–1789 and the next cohort comes later.» > «No Defranceschi baptism between 1799 and 1806.»

The sentence was written from memory of what Gologorica ought to show, and the search log four files away said the opposite.

⭐⭐ WHAT REPLACES IT IS BETTER, AND IT IS WHY THIS CORRECTION IS WORTH ITS SPACE. The two registers now say the same thing twice:

For roughly fifteen years the Defranceschi baptise no children in either parish, while standing at both fonts constantly, styled cives in both. That is a far stranger fact than a move between villages, and it was invisible while each book was read on its own.

⚠ THIS IS THE THIRD ERROR OF ONE SHAPE IN ONE DAY, AND THE SHAPE IS NOW THE LESSON. A correction that was itself a misreading; a «discovery» of confirmation rolls the archive had catalogued nine days earlier; and now an inference asserted about a book this archive had already read. All three were committed at the end of a long piece of work, in the sentence that ties the finding off. Added to the method: the closing sentence of a finding is where this archive is least careful and most quotable, and it should be the one line that gets checked against the search log before it is written.

· Corrected in place: /gologorica-line/ · /searched/

Earlier today this archive published a correction saying that a fast pass had misread the godmother of 16 October 1798, and that she was not Theresia but «Catharina Defranceschi». It went further: the Gologorica register of 1785 names a «Catharina Defranceschi, Cives ex Gollogorica», so this was announced as a second woman of the name crossing between the two parishes.

⚠ The correction was wrong. The first reading was right. Re-cropped at twenty times, with the mother's cell on the same line for alignment:

> «Joanns Salamon | Maria Salamon n. Picot | Antonius Picot (struck) X / «Theresia Defranceschi» X | civis | Franciscus Glogovaz Caplnus»

There is no Catharina at Gračišće. The 1785 Gologorica Catharina is untouched and remains exactly what she was; what is withdrawn is the bridge built between her and this entry.

⚠⚠ AND THE SHAPE OF THIS ONE IS THE POINT, BECAUSE IT IS NEW. This archive's rules are full of guards against publishing a bad reading — a hand is only legible against itself, never read rank off an abbreviation, crop past the cell. Every one of them was written for first readings. A correction is a claim like any other, and this one was made at speed, on the strength of a glance, precisely because correcting feels like the careful act. Added to the method as a rule of its own.

Also corrected in the same pass, and from the same cause. Four entries were dated 1802 and one 1801; re-read against the folio numbers and the mid-page year dividers they are 1801 and 1802. That dissolves a second thing this page announced: an apparent «disorder» in which M01070217 (1802) sat before M01070218 (1801). In folio order they run 1801, then 1801–02. There was no disorder. There was a misread header.

The twenty-one appearances themselves stand, and the parents' columns have now been read too — which is the finding this correction sits inside, and is unaffected by it.

· Corrected in place: /gologorica-line/ · /searched/ · /worklist/ · /corrections/

This archive concluded, and published in four places, that «the family was not established at Gračišće in the eighteenth century». It rested on a real and careful nil: the Pater column of the Gračišće baptism register read across thirty-six consecutive frames, M01070190–M01070226 — 1783 to May 1808, roughly nine hundred baptisms — with not one Defranceschi father.

⚠ That nil is TRUE and still stands. The conclusion drawn from it does not.

⭐ THE ARCHIVE HAD ALREADY FLAGGED THE GAP, IN WRITING, AND NOBODY CLOSED IT. The search record says it exactly: «the father column was read … a Defranceschi could stand as a mother's maiden name or a godparent, and one pair does in 1808. «None seen in twelve frames» is the claim; «none exists» is not.»

> The Patrini column of those thirty-seven frames has now been read, every row. It holds about twenty Defranceschi appearances.

WHO IS IN IT, AND THE CONDITIO COLUMN IS THE POINT.

Their condition is entered «CIVES» almost every time — townspeople of Gračišće — in a column whose other values are rustici, villici, agricolae, supani, judices. The register is not describing visitors. It is describing townspeople.

SO THE CORRECTED SENTENCE IS NARROWER AND MORE USEFUL: the Defranceschi were present at Gračišće, as cives, continuously from at least 1791 to 1808 — and in all that time had not one child baptised there. Present, in demand at the font, and not breeding in the parish. That is a far stranger and better fact than absence, and it is exactly the window in which the archive expected the family's arrival to become visible.

⚠ What is NOT claimed. Twenty sightings read at working magnification; three were re-read at six times and are certain. Every frame is named so any can be rechecked. ✔ Dated the same day: all twenty-one are now read against their date columns and run unbroken 1791 to 1808 but for 1803–05. ⚠ A later pass corrected the dates again, and withdrew a «Catharina» this archive briefly put into the table: the 16 October 1798 godmother is Theresia after all. No descent is traced from these people to the direct line here.

· Corrected in place: /errands/ · /searched/ · /worklist/ · /gologorica-line/

Verifying the Gračišće baptism book of 1667–1745 frame by frame turned up a contradiction inside this archive before it turned up anything in the register. Two pages state the result of the same reading of the same book, and they do not agree.

> /direct-line/ — «The Gallignana baptism book of 1667–1745 has now been read leaf by leaf … Across the whole of it the surname occurs exactly once. It is on folio 117, and it is not a household of this parish.» > > /open-questions/ — «The whole Gallignana baptism book of 1667–1745 has now been read leaf by leaf, and the surname occurs five times in it: this count in 1720; a Presbiter Philippus de Franceschi celebrating a baptism on 19 March 1734; and three times a Defranceschi of Gologorica standing godparent, in 1733, 1743 and 1744.»

Both say «read leaf by leaf». One says once. One says five.

They can be reconciled — the «once» is almost certainly once as the family of a baptised child, and the «five» counts every role including priest and godparent. But neither sentence says so, and a reader who meets either one alone is told something false about the other four. A count without its denominator is the same error this archive wrote a method note about yesterday, arriving from inside.

Fixed on the direct-line page, which now states what is being counted. The five-occurrence statement stands and is being verified frame by frame.

✔ And the entry itself is verified. Read independently at M01070065, folio 117: «Die 25 martij 1720 … Domini Comitis Damiani de Franceschis, ac Ursulæ eius coniugis», baptised by «Marino Josepho Bellaz, Præposito nec non Vicario Cathedralis Ecclæsiæ Petinensis». The archive's transcription matches the page word for word.

· Corrected in place: /direct-line/ · /open-questions/

This afternoon this archive published «⭐⭐ NASTAVLJA SE — the Gračišće baptism book stops in May 1808, and the film says so itself», as the headline of a commit.

Errand 35 already said it, in these words:

> «Its last written page, M01070226, is headed «Annus 1807»; the next frame is a printed card reading «NASTAVLJA SE» — to be continued. Everything after 1807 is on ZM34K-216.»

The errand also already carried the continuation's frame range (M01070239–M01070310), the fact that the family holds ZM34K-214 and 215 from the same series, a drafted letter in Croatian, and the finding that there is no online route — tested twice by two methods.

⚠⚠ THIS IS THE FIFTH INSTANCE IN THREE DAYS, after Francisca, the priests' origins, Joanna's house 4 and the house 5 page. Four method notes have been written about it. The notes are not the problem.

The specific failure this time is nameable: the grep that has been adopted runs on a person's name. This finding had no person in it — it was about a frame. So the habit did not fire. The rule is widened: before publishing anything found on a frame, grep the archive for the FRAME NUMBER too — M01070226 would have returned errand 35 in one second.

What survives from that commit, and it is most of it: the Pater column swept across thirty-six frames with no Defranceschi father; the frame-map correction; the «Fede di Nascita» certificate and its Trieste–Capodistria letterhead; the water-damage label on the prose half. The headline was not new. The body largely was.

· Corrected in place: /gologorica-line/ · /errands/

Yesterday this page announced «⭐⭐ Both brides came from Gračišće — houses 4 and 5 — and the census said so in a column I never read», closed work-list row 184 on it, and raised two more rows to carry it forward.

It was documented in this archive on 30 September, on Joanna's own roster row, from the same frame:

> «The 1826 status animarum of Gologorica records her in house 3 … born 25 June 1818 … They married on 13 February 1843, and the provenance column records her coming «Ex Parochia Gallignanæ sub N.º 4» — out of Gračišće house 4.» > > «Her marriage is one of two in three years that took a Defranceschi daughter out of Gračišće and into Gologorica — hers from house 4 in 1843, and Francisca's from house 5 in 1840.»

Six of her children are listed there too. The «column I never read» had been read, and the conclusion drawn from it — two brides, two adjacent houses, three years apart — had been written out in a sentence.

⚠⚠ THIS IS THE THIRD TIME IN TWO DAYS. Francisca's origin was announced as new yesterday and withdrawn the same day; the priests' origins were half written up before a grep stopped them; this is the same error again, and the second method note about it was written yesterday.

So the note is not the problem. Running it is. The check that would have caught all three is one command on the person's name against roster.json before the finding is written, not after it is published — and from now on it is the first thing done when a name is read off a page, not the last.

What stands: the structural point that the «Locus ex quo vel ad quem» column exists and had never been harvested is real, and harvesting it produced the migration table, which is new. The two brides are not.

· Corrected in place: /gologorica-line/ · Joanna Defranceschi (Sestan) · /worklist/

Yesterday this archive announced as new that «Francisca uxor n. Defranceschi, b. 3 Februarii 1819» had married «Antonius Defranceschi» — a marriage inside the surname — and that it «has not shown the surname marrying itself at Gologorica before». It then raised two work-list rows to go and find who she was and where the other Defranceschi family lived, and published that there was such a family «that this archive cannot yet name».

All of that was already here, with its sources, from 30 September.

> This archive's own record for her, written a week ago: «She came from the parish of Gallignana (Gračišće), house 5, on her marriage of 29 January 1840 — «Ex Parochia Gallignana sub N.o 5», 1826 status animarum, M01069328» — and the marriage-inside-the-name was settled the same day «beyond doubt within the same book, on three constructions in one hand».

The frame it cites is the frame this walk had open.

⚠⚠ AND THE CAUSE IS WORSE THAN THE ERROR. An opening of this census is one table across two pages: the left half carries names and birth dates, the right half carries marriage, «Locus ex quo vel ad quem» and vaccination. This walk read the left half, wrote in its own search log that the right half was «not read», and then built four findings and three work-list rows on the gap it had just declared. A column recorded as unread is a blank in the reading, not a blank in the record, and nothing may be concluded from its silence.

⚠ This is the second time this exact failure is on the record. A method note of 21 September already says that with thirteen hundred people on the books, the first question about a find is whether it is already here. It was not asked. The note is restated with teeth.

What survives, and it is genuinely new: reading that column has now produced «Ex Parochia Gallignanæ sub N.ᵒ 4» for «Joanna uxor nata Defranceschi», with her marriage of 13 February 1843 — which answers the question reopened two days ago. Her house 4 and her date are the finding. Francisca's house 5 is not.

· Corrected in place: /gologorica-line/ · /worklist/

Yesterday this archive closed worklist 184 — why a Defranceschi died at house 6 in 1858 and 1874, in a house that was Deprato in 1801, 1809 and 1817 — by writing that the death book's index puts two household blocks at that number and so there was «no family connection at all». It is withdrawn. The 1826 Status Animarum refutes it in two lines.

> «Vincentius Sestan, cives» · «Maria Uxor nata DEPRATO» (M01069324) > «Antonius Sestan defᵗⁱ Vincentii et Mariae» · «Joanna uxor nata DEFRANCESCHI» (M01069323)

Vincentius Sestan's wife was born a Deprato; their son's wife was born a Defranceschi. Two marriages, two generations, and they are the exact three surnames the index stacks at house 6. The house did not hold strangers in succession — it moved with a marriage.

What stands: the index does give six house numbers two blocks each, and a house number is an address rather than a household key. Those are reads of the index and nothing here touches them. What falls: the inference drawn from them about these families.

⚠ AND THIS IS THE SECOND WRONG ANSWER TO ONE QUESTION, FROM THE SAME BOOK. The first claimed Mathaea Defranceschi was born a Defranceschi and married a Deprato, on a reading of how the register names married women that was exactly backwards. Both answers were deductions about kinship made from a book that does not record kinship. A death register's index records households and addresses. The Status Animarum writes the relationship into the name — «uxor nata Deprato», «mater nata Bembo», «Mariana ex fratre nepos» — and this archive had walked that book six days ago and asked the other one twice.

Worklist 184 is reopened, not re-closed. The method note is here.

· Corrected in place: /gologorica-line/ · /worklist/

This archive published that Josephus Lambertus Defranceschi, born 13 March 1808, died on 22 December 1857. He died on 22 December 1878.

> «1878 Decembris 22. · 7 · Josephus Defranceschi, possidens, munitus SS. et sepultus die 24. Dec. 1878 · 70 · Senectus» (M01069295, p. 132)

Seventy on 22 December 1878 closes on the declared birth — seventy years and nine months. At 22 December 1857 he would have been forty-nine.

⚠ AND THIS IS THE SECOND TODAY, WHICH IS THE REAL POINT. Clementina Maria Defranceschi was corrected this morning from 10 March 1840 to 10 March 1870*.

In both, the day and the month were already right. Only the decade was wrong. That is the signature of a transcription slip, not of two different people being confused — which is what both rows had worried about in exactly those words: «a gap this wide means one side is describing a different person's death».

⭐ And both rows had said the remedy was the burial register. It was walked today, and it answered both. In both cases the family tree was right and this archive was wrong.

⚠ Exactly three rows in this archive carry that «disagree about the DEATH year» note. Two are now closed; the third — Lina Defranceschi, tree 1923 against 1925 — is still open and should be read with the expectation that the tree is right.

Fixed: death 1857 → 1878 on the roster row and @I36@, re-declared in data/decisions.json in the same commit.

· Corrected in place: /gologorica-line/ · Josephus Lambertus Defranceschi

This archive published that Clementina Maria Defranceschi, born 3 October 1822, died on 10 March 1840. She died on 10 March 1870. The day and the month were right; the decade was not.

⭐ The good part: the row had already said how to settle it. A note of 20 September 2026 recorded that the family tree said 1870 and this archive said 1840, that «a gap this wide means one side is describing a different person's death», and that «this needs the burial register, not a judgement». The burial register was walked today and answers it:

> «1870 Martii 10ma · 7 · Clementina Defranceschi, possidens, munita SS Sacram. et sepulta 11. Martii · 47 · Phthisis» (M01069286, p. 114)

Forty-seven on 10 March 1870 closes exactly on the declared birth of 3 October 1822. At 10 March 1840 she would have been seventeen. The tree was right and this archive was wrong.

And a second pointer had been sitting on the same row since 5 October: a Clementina Defranceschi stands godmother in 1841 — which a woman dead in 1840 cannot do.

Fixed: death 1840 → 1870 on the roster row and on @I40@, re-declared in data/decisions.json in the same commit. The pid keeps 1822 because it encodes the birth year, which is unchanged.

⚠ What this says about the practice. The error was flagged, dated, and assigned a specific remedy three weeks before it was fixed, and the remedy worked exactly as written. That is the system behaving correctly — but it also means the wrong year was live on the site for those three weeks, which is why the worklist's open questions are worth working through rather than accumulating.

· Corrected in place: /gologorica-line/ · Clementina Maria Defranceschi

Published this afternoon: that Mathaea Defranceschi, dead at house 6 in 1858, was born a Defranceschi who had married a Deprato — and therefore that the two «De—» families this archive has spent weeks separating were joined by marriage.

The argument rested on one premise: «these registers enter a married woman under her birth surname as a matter of course». They do the opposite. They use the HUSBAND's surname, sometimes appending «nata X».

What disproved it, eight pages further on in the same book: «1863 Julii 16 · 7 · Lambertina DEFRANCESCHI, civis · 77 · senectus». Lambertina was born Peschle, of Volosca, wife of Josephus Defranceschi and mother of Carlo De Franceschi. The register gives her husband's name. And the premise's own star witness says the same thing read properly: «Catharina DEPRATO nata Rovisi … uxor Joh: Baptista» — Deprato is the husband's family; Rovisi is hers. I had cited that very entry as evidence for the opposite.*

Corrected reading: «Mathaea Defranceschi» was the wife of a Defranceschi — which is what the line record already said, «Wife of Leonardus (house 6)» — and her birth surname remains unknown. The genuinely open question is now sharper: why does a Defranceschi wife die in the house that was demonstrably Deprato in 1801, 1809 and 1817?

⚠ Two things about how this happened. First, it was caught by continuing to walk, not by any check — nothing in the gate can test a premise about a register's habits. Second, this archive has made this exact mistake before: a sentence about what a book does from a date is a claim about every page after it, and so is a sentence about how a book names people. The claim stood about two hours. It is withdrawn here rather than edited away.

· Corrected in place: /gologorica-line/ · «The two «De—» families may not be two families after all»

Writing up a new finding about duplicate FRAMES on the microfilm, I gave it the key duplicates in gologorica.json. That key was already in use. It held the 1836 collation certificate — «Quae copiae collationatae anni 1835 concordant cum hoc originali libro… Petinae die 7 Maij 1836» — the finding that a second, certified set of these registers left the parish every year, which was itself only recovered after the crop-truncation error was caught. A json.dump of a dict with an existing key does not warn. The block was gone, with its six paragraphs and its plate.

How it surfaced, and it was luck rather than a check. The next section I tried to add asserted that its anchor appeared exactly once in the page and the assertion failed — because the page now rendered g.duplicates twice, the original section and the one I had added. An assertion written for a different purpose caught it. Nothing in the gate would have: the data was valid JSON, the page built, every link resolved, and check-story counted the same number of findings, because the count does not know which finding.

Fixed. The original block is restored byte-identical to its previous version (verified against git), every other key confirmed present, and the new finding moved to its own key dupframes. Both now render once each.

⚠ The lesson, which is now in method: before writing a new key into a shared data file, check whether it is taken. A one-line assert key not in data costs nothing. This archive already knew that ids must be minted once and frozen — it had not extended the thought from person ids to section keys, which are just as published and just as easy to clobber.

· Corrected in place: /gologorica-line/ · the block «A second set of these registers left the parish every year»

was That the Search Register listed every source this archive had examined — «197 sources examined», as its own header said.

now It was showing 112 of them. Eighty-five rows were on the page's data file and not on the page. The cause is dull and the consequence is not. The template grouped rows by an outcome field and knew five values — yield, partial, nothing, blocked, pending. Over a year the rows had been written with twelve: 70 of them say «hit», seven say «miss», and the rest are «found», «note», «mixed», «correction», «clean». Anything the template did not recognise was silently dropped. No error, no gap, no warning — the row simply was not drawn. So the register understated the work by a factor approaching two, and did it in the one place on this site whose entire purpose is to say honestly how much has been looked at. A page that exists to record negative results was itself losing results. And three of the eighty-five were written today, by me. The Hungaricana searches and the ANNO search were all filed as «found», published, and were never visible. I found them missing only because I went back to check that a published finding had actually appeared — which is a habit this archive should make routine rather than occasional. Fixed two ways: the twelve words are now mapped onto the five outcomes, and any value that still is not recognised falls into an «Other» group rather than disappearing. It is no longer possible for a row to be invisible. 197 render.

· Corrected in place: /searched/

was Twenty-four of this archive's 112 households named a De Franceschi woman as the father and her husband as the mother. «father: Michiela De Franceschi · mother: Domenico Giachin». Forty people's roster entries said things like «Domenico Giachin — named as mother on 2 baptisms».

now Every one is inverted. The women are the mothers; the men are the fathers. All 24 are De Franceschi daughters married out — which is why the error clusters exactly where it does.

how The households built from the FamilySearch index take the two parent names the index returns and assign them by position, not by sex. Where the De Franceschi parent was listed first — which is what happens when the mother is the one carrying this archive's surname — the pair went in backwards. So the bug bit precisely the people this archive is worst at tracking. The whole Vodnjan argument is that «the surname keeps its sons and loses its daughters»; twenty-four married-out daughters were in the data and the data had them down as men. Found while testing something else entirely — chasing whether one Michiela De Franceschi fu Giuseppe was a daughter of the Giuseppe this archive already documents. The check that found it is now in scripts/audit.py and runs on every publish: a female given name in the father field. It would have caught this the day it was introduced.

· Corrected in place: /households/ · /people/

was That FamilySearch's full-text search returned 1,952 results for Gologorica, 1,719 for Gračišće and 1,885 for Defranceschi — and, noting how implausibly alike those three numbers were, that they «have not been tested» and should not be trusted.

now Every one of the three is wrong, and the real figures are not close. Gologorica returns 19. Gračišće three pages. Crikvenica fifteen. Ližnjan two. Defranceschi returns about 9,900 — five times what was published, not a fifth of it. And Vodnjan returns nothing at all: the Croatian name post-dates every record in the corpus, which uses Dignano — about 12,200 of those, most of them the other Dignano, in Friuli.

how The number was scraped with a loose pattern — ([0-9,]+)\s*Results — off a page whose visible text ends «… Border Crossing Records 1924-1952\nResults per page:». It matched a year out of a collection title. «1,952 results» was a border-crossing collection's end date standing next to the words «Results per page». What makes this worse than a scraping bug: the page said the numbers looked wrong and published them anyway. Three unrelated searches returning numbers of the same size was called «the shape of a number that counts something other than what you think» — and then the one check that would have settled it, look at how many results are actually on the page, was not run. It takes one request. Flagging a doubt is not the same as resolving it. A figure you have publicly called untrustworthy should not be in the sentence at all. The finding underneath is unharmed and is better than it looked: 19 hits for Gologorica is a set small enough to read end to end, which is exactly what makes the village sweep worth doing.

· Corrected in place: /method/

was Nothing — that is the problem. The account of Paputchka's death, some five hundred words, silently vanished from the transcription.

now The scanner destroyed its opening sentence, returning «WS CAPES FUG PS PEPE AU SII Pony FRU. OI WC PI PO UT UU TET» where she had written «We experienced "true" love with our Siberian Husky Paputchka». A rule I wrote to strip scanner noise matched that line — and because prose is stored one line per paragraph, deleting the line deleted the entire paragraph.

how Restored word for word from the original, supplied by David, and marked on the page as restored so no reader has to wonder which text they are looking at. The lesson is specific and unpleasant: a cleanup rule that operates on a whole paragraph can destroy a whole paragraph. Noise-stripping must be scoped to the noise — a few tokens — and never to the container it sits in. It was also found by the reader, not by me: nothing in the build warned that five hundred words had gone, because nothing was counting. A reconciliation was then run over the whole page — every substantial line of the raw scan tested against the published text — and it found four more losses of the same kind: the closing paragraph of the husky piece, which had been left beginning mid-sentence at «anger and rage»; a sentence from A Suicide Avenged; a line of dialogue from it; and part of the godparents letter. All four are restored and marked. One passage is unrecoverable — three lines describing a person, from the 2004 course — and stands on the page as a stated gap rather than a silent one.

· Corrected in place: /cheryl-lerena/

was Her own introduction to the first poem was published word for word, and it named the man she was then married to in a sentence describing «a long standing affair, committing adultery for years».

now That should never have gone up. It is an allegation of adultery against a named living person, published by this archive. That the words are hers and that she published them herself does not transfer the risk or the discretion — the archive chose to reproduce them.

how The name is removed; the sentence now reads «my marriage was over». It is the single alteration to her text anywhere on that page, and the page says so. The failure was not ignorance of the rule. The same rule had been applied an hour earlier, and correctly, to a different document — a statement of allegations that was excluded entirely for naming living people. Then a transcription was published verbatim without running the same test over it. A rule applied once is not a rule; it has to be run against every page that goes out.

· Corrected in place: /cheryl-lerena/

was Two pages give her birth as 8 March 1926. A third page quotes the certified extract of the Senj birth register, attached to her son's 1951 entry, which gives 3 August 1926.

now Both have stood on this site at once and nobody noticed. Written as numbers they are 8/3/1926 and 3/8/1926 — a day and a month exchanged, almost certainly where a European date met an English-speaking translator. This archive does not know which is right.

how Neither has been quietly deleted. Both now stand, marked as in conflict, and the Senj birth register for 1926 is an errand. The lesson is one this project keeps relearning: a date that appears twice in an archive should be diffed, not trusted twice.

· Corrected in place: /senj-line/ · /crikvenica-house/

was Anton Rudolf Defranceski, born 20 January 1890 at Klenovica house 22, died in 1943 — recorded here «per the family». The same page also has him on the Pfullingen list of foreigners in 1947 and certified by that town's mayor in 1950.

now Those cannot both be about the same living man. Either the 1943 death is wrong — and it was only ever family testimony, never a register — or the Pfullingen documents are about a different Anton. The archive printed both and did not notice.

how Surfaced only because the family's oral history of 1948 made the widow's situation matter. Written up on Nineteen forty-eight, and the Crikvenica death register for 1943 is now an errand. Lore is worth collecting partly because it makes you re-read your own pages.

· Corrected in place: /crikvenica-house/ · /direct-line/

was That FamilySearch's tile server had blocked the death-register reading, recorded as blocked in the Search Register and written up on Method as a second instance of when the service says stop, stop.

now The tile path was wrong. Tiles live at …/image.xml's sibling image_files/{level}/{col}_{row}.jpg, and the reader was asking for …/{level}/{col}_{row}.jpg — no image_files. Today that same wrong path returns a plain 404; yesterday the security service answered the unknown path with a 403. So the reading was stopped by my own broken URL, dressed up as somebody else's refusal. A real block did come later, after several hundred tile requests, on image.xml itself — and that one is recorded honestly.

how Check that the request is well-formed before concluding you have been refused. A 403 on a path that does not exist is not a verdict on your conduct. The cheap test — does the correct URL work? — costs one request and was not run.

· Corrected in place: /searched/ · /method/

was A hand-tinted portrait of a bearded man in a bow tie, published as Josip Defranceschi, born at Gračišće 22 March 1863 — David's great-grandfather, and the only likeness the direct line had.

now It is from the Imotski branch, and the sitter is a Josip of that family. No likeness of the Crikvenica Josip is known to this archive. The companion portrait published as his wife Ana is withdrawn too, because its only identification was that it matched the man beside it.

how Two branches' photographs sit in one family collection. A card carried the name, this line has a man of that name, and this archive matched the face to the name without asking which Josip — while checking every register plate on the same page against Pazin and Ledenice. The portrait was never checked against anything. Identified by David, who holds the originals. The retired 2018 GEDCOM spliced Imotski onto this line as well: the same seam has caught this archive twice.

· Corrected in place: /crikvenica-house/

was That the seventy-eight Vodnjan records were really Vižinada's, because resolving them to film 005497893 and asking the catalogue who made that film returned «Rimokatolička crkva. Župa Vižinada».

now A reel can carry more than one parish. That one holds Vižinada's deaths 1754–1831 and Vodnjan's baptisms 1815–1866 (koha:963267). The records are Vodnjan's.

how The catalogue answers a film-number query with the first creator on the reel. The method was sound and the question put to it was wrong. It stood on the site for about an hour, published within twenty minutes of being reached and without a single image being opened. What made it look right: the one death on that reel really is Vižinada's — it is in the Liber Defunctorum of Castellier, Kaštelir.

· Corrected in place: /fiume-1626/

was Eight reconstructed Vodnjan households and their children carried Vižinada as their parish in the roster, and the households themselves had no place at all.

now They are Vodnjan. Eight households placed, forty-four rows moved, and Vodnjan moved out of unplaced into Southern Istria.

how The correction above was published on the narrative page and never reached the data. David asked why the headcount had not moved all day; it had not, because the roster was three commits behind the research and still carried the error the essay had already retracted. Fixed, and the rebuild order is now written into the queue notes.

· Corrected in place: /people/

was The Imotski family joined to David's own line, in the 2018 GEDCOM this archive inherited.

now They are separate. The GEDCOM is retired — 8,286 people against the live tree's 15,643, and David's own verdict: «really old and really wrong».

how An old export was trusted as a source. It is now used for nothing. Note how often Imotski turns up in this list.

· Corrected in place: /imotski/

was Per-place record counts for this surname across sixty Istrian and Kvarner towns, and a research queue carrying «Vižinada — 53 records, unread», «Draguč — 25», «Dubrovnik — 38».

now The place parameter on this search endpoint does not filter. Ten values were tested, including «Zzzznowhere» and «Atlantis», and every one returned the same total: 3,652. Vižinada and Draguč return identical result sets. Dubrovnik returns the unfiltered list. The queued «Vižinada 53» records are on the Vodnjan film and have already been read.

how The parameter is a relevance boost dressed as a filter — no error, no warning, just the whole set back in a different order. The sweep survives because its method independently checked each record's own place text before counting it; the queue did not, and carried three phantom tasks for days. The reliable route is to resolve records to their film and read the catalogue's item list — remembering that a film can carry more than one parish.

· Corrected in place: /istria-sweep/

was That forty-five per cent of the indexed records carry no town, that the film number in each record is the way to a parish, and — presented as the day's lesson — that a reel can carry more than one parish, so the film is «a fast first pass and not an answer».

now All three were already on this site. The parish-map section of the same page says the index «usually gives no parish… every persona carries the ark of the image it was read from, the service turns that ark into a DGS film number, and the catalogue names the film's parish». And the correction entry above it, from 9 September, already ends: «The reliable route is to resolve records to their film and read the catalogue's item list — remembering that a film can carry more than one parish.» The 45% is a measurement of something this archive had already stated qualitatively. It was written up as though nobody here had noticed.

how By not reading the page being added to. Four sections were appended to /istria-sweep/ in one afternoon without the existing parish-map section being read first — the same page, eighty lines further down. What is actually new, and it is narrower: q.fatherSurname and q.motherSurname are honoured when either is the sole name term, which the reference note said was impossible and which took the pool from 954 records to 1,619; and resolving a record through its image ark to the book waypoint is finer than film-to-catalogue — it settles a two-parish reel automatically instead of requiring somebody to remember to check. The measurement that 110 of 112 records on film 005497893 are the second parish's is new. The idea is not.

· Corrected in place: /istria-sweep/#noplace

was That an Aloÿsius Defranceschi appears at Albona in 1615 — «sixty-four years before the Mione headmen's list, and eighty-five before the date Carlo gives for his family's arrival». Carried as one of three key findings of the September sweep, pinned on the atlas, and used as the outer bound of the family's presence in Istria.

now Three later measurements cannot find it. The rebuilt sweep on the same page — written precisely because the first method matched a truncated field and stopped early — gives Labin 1789–1892. The by-image test of 13 September gives 1826–1866, and six of Labin's fourteen indexed hits are Franceschich, Francescich, Francocich, which are not this family. And this archive's own register, 1,508 rows each carrying the id the service returned for it, holds no 1615 Labin row; its only 1615 row is an Andriana Franchini. So the 1615 exists in one place only: the table built by the method this page has already declared broken. And the source of the number has since been found. The only Aloÿsius Defranceschi in any harvest on disk is born 1790, son of Mathæus Defranceschi and Osvalda Labinaz or Dominj — a Vižinada household. The Gasparo De Franceschi named beside him is in none of the data. And Labin's filmed volume list contains a book called «Deaths (Umrli) 1615, 1585-1649». So the entry was a man from the wrong parish dated two centuries early, a man who does not exist, and a date taken from a shelf-list instead of a record. It is withdrawn, not merely disputed. What stands instead: Rovinj 1619, then Svetvinčenat 1628. Both are in the register, both have an image behind them, and the second was read off the page the same morning.

how By checking a load-bearing number against the archive's own data instead of against the page that first published it. The check cost one grep. It was prompted by writing a new place page for Labin and finding nothing in the register to put on it. Withdrawn the same day after searching every harvest file for the name rather than only the published tables. That search cost one grep too, and should have been the first thing done rather than the second.

· Corrected in place: /istria-sweep/ · /map/

was A birth year for each of ten married-in women — «b. 1822» for Lucia Bonassin, «b. 1816» for Maria Malusá, «b. 1841» for Maria Dorliguzzo, and seven more.

now None of those is a birth year. Each woman is the mother of a household whose first child is baptised in that very year, or in one of the years she was already bearing children. Lucia Bonassin «b. 1822» is the mother in «Gio Maria De Franceschi × Lucia Bonassin, 1822–1831». Maria Malusá «b. 1816» is the mother in «Pietro De Franceschi × Maria Malusá, 1816–1824». Anna Defranceschi (Voran) «b. 1865» is the mother in a household running 1857–1865. The year is the first time she appears in the record — the baptism of her eldest known child. It was carried into the roster as a birth. Nobody is born in the year their own child is born. All ten now show the year as a first appearance, with the birth recorded as unknown.

how Found by asking why the pedigree gate was suppressing fifty-five charts. The gate was doing its job: it refuses a parent who would be under twelve at the first birth, and these women were failing that test because their data said they were nought. The charts were the symptom; the birth years were the fault. And one rule was tried and withdrawn inside the same hour. Generalising it — any person whose recorded birth is less than twelve years before a household they parent — «corrected» 113 rows, including an Antonio Defranceschi born 1897 attached to an 1825 household and a Josip Defranceski born 1947 attached to one from 1890. Those are not bad birth years; they are name collisions, which this archive proved the same morning when «Maria Defranceschi» turned out to be thirteen women in thirteen parishes. The rule assumed the match was right and the year wrong; in 103 of 113 cases it was the other way round. Reverted, and narrowed to the one case where the name form itself identifies a wife — «Maiden (Married)» — which no father and son can share.

· Corrected in place: /who/ · /people/

was That the Šterna household ran Gio Batta De Franceschi × Zanetta Gravisi, and that their son Bortolo's marriage stood behind two entries in the book of 1723–1820. Published on the morning of 13 September, on the strength of the index's four names.

now Zanetta Gravisi is Bortolo's wife, not his mother. The index lists groom, groom's father, bride and bride's father; this archive read four names as two couples. The page reads: «Bortolo figlio del q. Gio: Batta de Franceschi dalla Cargna… ha contratto legitimo Matrimonio coll'Ill.ma Sig.a Zanetta Gravisi figlia dell'Ill.mo Sig.r Marchese Nazario della Villa di Cepich di questa Pieve», 12 November 1777. And there was never a second marriage. The other «record» is a modern archivist's pencil finding-aid slip filmed with the book, which the index counted as a separate entry — and which independently confirms every word of the reading.

how By opening the two images instead of reasoning about the index rows, which is what the same page had already said was the only reliable move. The correction was published the same day as the claim. It is also the most valuable correction this archive has made, because going to the page produced the sentence the whole project has been missing: an Istrian De Franceschi whose origin is written down.

· Corrected in place: /places/sterna/ · /istria-sweep/

was That the largest of the married-out daughters had nine children by an Antonius Cerneca between 1831 and 1846. Published this morning as the head of the lost-daughters queue.

now Her husband is Joannes Bergamo, and they are not children. The Svetvinčenat marriage book after 1815 is a ruled table with four printed column headings — SPONSI, SPONSAE, GENITORES SPONSI, GENITORES SPONSAE — and the index flattens all four parents into one list. «Vincentius Bergamo, son of Joannes Bergamo and Francisca Defranceschi, marries Euphemia Cerneca, daughter of Antonius Cerneca and Joanna Ferrara» arrives as one row of five names, and this archive read names three and four as the husband. The five rows are the weddings of four sons — Joannes 1831, Dominicus 1836, Vincentius 1840, Antonius 1846 — not nine baptisms. She remains a married-out daughter; that part of the queue's logic is unaffected. Every husband in this archive inferred from an o list on a post-1815 marriage register is now suspect.

how By opening the page. The column headings are printed in capitals across the top of every opening and settle it in one look. The same morning's own text had already said that the reliable move is to read the register rather than the index; it took until the afternoon to apply that to a row this archive had itself put at the top of a queue. Bergamo is also the point. Kandler lists the Carnian families at Sanvincenti as Defranceschi, Lupieri, Fiorencis, Fabris, Crosilla and Bergamo — so this is two Carnian houses intermarrying, which the wrong husband had hidden.

· Corrected in place: /register/#resolved

was An ancestor chart published a few hours earlier, in which Hedviga Blažević was drawn as Ivan Defranceski's mother, Uršula Kovačina as Anton Rudolf's mother, Anna Kalanj as Josip's mother, Maria Fernasar as Antonio's, and Francisca Salomon as Giovanni Battista's.

now Each of those women is the wife of the man she was drawn under, not his mother. The whole maternal side of the chart was shifted by one generation. A wife belongs above her children, not above her husband — Anna Kalanj is Anton Rudolf's mother, not Josip's; Hedviga Blažević is the mother of the living generation, not of the man she married. Rebuilt so that each couple's children sit below them: I. Defranceski ← Ivan (1925–1995) and Hedviga Blažević; Ivan ← Anton Rudolf and Uršula Kovačina; Anton Rudolf ← Josip and Anna Kalanj; Josip ← Antonio and Maria Fernasar; Antonio ← Giovanni Battista and Francisca Salomon. Giovanni Battista's own mother is now a wall, because she is genuinely not named anywhere.

how By printing the tree as a list of [role] pairs instead of looking at the rendered chart. On the screen the error is nearly invisible — every box is in a plausible place and every name is right. On one line of text per person, with «father» or «mother» against each, it is obvious in a second. The lesson is the same one this archive keeps relearning in a different costume: a shape that looks right is not a check. The descent view built from the same data was correct throughout, which is why the error survived — the two views disagreed and nobody had asked them to agree.

· Corrected in place: /tree/

was Published this morning: «985 more volumes now open at page one», counting 343 Dalmatian and 642 Carnian. The Carnian half was described exactly like the Croatian half.

now The images of the entire Udine collection are restricted. The viewer answers «Image Restricted — access is typically determined by local laws or the custodian». Tested at Colza, Comeglians and Paluzza; all three are shut. The Croatian volumes — Istria, the Kvarner and Dalmatia — open normally, which is how the mistake survived a spot-check: the one volume tested end to end that morning was Tinjan, and Tinjan is Croatian. So the honest count is 652 volumes that open and 642 that are addressable but gated. The link is still worth having — it names the exact volume and its register number, which is what a reading-room request needs — but it is a catalogue entry, not a document.

how Found within hours, by going to do the thing the page promised: reading Colza to settle whether Ovaro's registers are filed under it. The page said the volume was open; the volume was shut. A link was tested for the wrong country and the claim was written for both.

· Corrected in place: /parish-books/

was Published this evening, as the headline of the 23andMe section: that David's 0.2% Southern Indian & Sri Lankan is «1/512 — nine generations, about 1750–1780», that this is «the window this page reasoned to from the Y chromosome alone, before the trace was known», and that «two instruments, two companies, one century… is a convergence, and it is the strongest thing on this page.»

now There is no convergence. The trace is maternal. Both of David's parents share their 23andMe results with him, and the comparison panels say it plainly: Ivan 0.0% Central & South Asian. Cheryl 0.5%. David 0.2% — half his mother's, as inheritance requires. The Sub-Saharan African behaves the same way: Cheryl 0.8%, Ivan 0.0%, David 0.3%. The trace comes down the mother's line. L-M27 comes down the father's. They have nothing to do with one another, and the arithmetic that joined them joined two unrelated numbers. The consequence is not neutral — it makes the Y finding starker. Ivan is 99.9% European with no non-European trace at all, and he carries L-M27 on two companies' chips. The haplogroup now has no autosomal corroboration whatever.

how Caught the same evening, by reading the parents' kits. The failure was not the reading — it was the publishing. The same article said, three paragraphs below the claim, «Which parent the trace came from is not yet known… Ivan's own 23andMe kit will answer it free» — and then ran the convergence as the headline anyway. A stated caveat is not permission to lean on the conclusion. The check took two clicks inside an account this archive was already signed into.

· Corrected in place: /dna/#ydna23

was Mione's panel showed one record type — Marriages, and thirty-three volumes filed under «Other: titles this archive cannot classify». Twenty-three other Carnian and Istrian comuni said the same. The map could not draw a births or deaths timeline for the parish this whole family comes out of, and could not show that Mione's civil births stop in 1810 — a finding the archive had already published in words on the Carnia page.

now «Nati» and «Morti» were unmatchable. The patch of 15 September that taught the classifier its Hungarian words wrote the word boundaries into a non-raw Python string, so «\\b» was stored as an actual backspace character. The pattern became «nati⌫» and «morti⌫» — two alternatives that can never match anything, in a file where nothing looks wrong, because a terminal prints a backspace by moving the cursor. 248 volumes across 24 places were in «Other» because of it, sixteen of them already read. Mione now shows Births 1806–1810, ALL READ, Marriages and Deaths, each with 1807 named as a gap. Found because David asked whether the green on one Gračišće row meant the same as the green on the block around it. It did not — that was a second bug, fixed the same hour — but checking the same thing on Mione and Ovaro is what turned this one up.

· Corrected in place: /map/ · /carnia/

was Prato — sixty-two volumes of the Val Pesarina — sat on the map at 43.88 N, 11.10 E: Prato in Tuscany, three hundred kilometres from Carnia. Priola sat in Piedmont. And Pesariis (32 volumes), Colza (25), Pieria (10) and Osais (8) had no dot at all, so their panels could not be opened and 75 volumes of the origin valley were invisible.

now Three faults, one family of causes — a name looked up without a region. The four missing villages are in GeoNames under compound names: Osais-Pesariis, Tartinis-Colza, Pieria-Prato Carnico. Only whole strings were indexed. The parts are indexed now, always below a place that carries the name on its own. Prato and Priola are real names in more than one province, and the index keeps one row per name, ranked by population — so Tuscany's Prato, with 190,000 people, beat Prato Carnico's 1,065. A place catalogued only by a provider that works one valley is now resolved inside that valley: the Udine sources get a Friuli box, and the candidate in the box wins. «Prato» is written down as Prato Carnico rather than left to a lookup to guess again. It is what the collection means — its own frazioni, Pesariis, Osais and Pieria, are listed beside it in the same collection. Found because David asked to check the rest of the Carnia panels after the «Nati» and «Morti» fix. Six of the thirty were wrong or missing.

· Corrected in place: /map/ · /carnia/

was Vodnjan stood on the map twice — one marker with nine volumes, and another a few pixels away with a hundred and thirty-three. Umag and Motovun were doubled the same way. A reader who clicked the nearer dot saw a ninth of the shelf.

now Pazin's register list writes a town under both its names at once — «Vodnjan / Dignano», «Umag / Umago», «Motovun / Montona» — and the whole string was being keyed as a third place. Three towns and 383 volumes were split off from themselves. The town is the first name now and the second is filed as what it is also written as, which is a thing this panel already knew how to say. Vodnjan is one marker with 142 volumes — 41 births from 1559, 44 marriages, 41 deaths, two real gaps named, and fifteen volumes that exist on paper and nowhere else.

· Corrected in place: /map/ · /vodnjan/

was Ledenice — where Anton Defranceschi was born in 1890 — was pinned at 48.93 N, 14.62 E: Ledenice in South Bohemia, eight hundred kilometres from the Vinodol. Ragusa, which is what every Venetian document calls Dubrovnik, was pinned in Sicily.

now The gazetteer keeps the most populous bearer of a name, which is right almost always and wrong exactly here: Czech Ledenice has 2,000 people to the Vinodol village's 164, and Ragusa in Sicily has 69,000. Population cannot tell a namesake from a home town. Both are now forced to the country the archive means, written down in the builder with the reason beside them rather than left to a count of inhabitants. Ledenice is at 45.14, 14.84; Ragusa resolves to Dubrovnik. Found while checking the Istrian panels after the Carnia ones.

· Corrected in place: /map/

was The Findmypast sweep was published with a finding headed «A Puerto Rican cluster of the name» — fifty-two records across eleven towns, offered as a lead on the surname this archive is built on.

now Every one of the fifty-two is a woman, and reading the names in Spanish says why. Filomena Franceschi Dávila de Padilla. Carmen Callejo y Ferrer de Franceschi. «de» is the marital particle there, not the Italian nobiliary one. A woman who marries a Franceschi is indexed de Franceschi; the men of the family are filed under plain Franceschi and a search for «de Franceschi» never sees them. The cluster is real. It is a Franceschi family, seen through a keyhole, and the surname in it is not this archive's surname. Corrected on the page in place, with what the right search would be. Found within the hour, while building the map layer from the same 644 records — the place data made the pattern visible in a way the summary count never could.

· Corrected in place: /findmypast/

was Carnia and Istria were checked because David asked about a panel, and nine places came back wrong. Nobody had looked at the other twelve hundred.

now A script now applies the two tests the builders already had. A place catalogued only by a provider that works one region cannot be outside that region, and a place whose own name carries a country cannot be in a different one. It found six more. Senja, Sv. Juraj — the Senj deanery's coastal parish, eight kilometres from the town this family's grandmother was born in — was outside Belgrade. Sv. Jakov-Šiljevica, which is the old name of Jadranovo four kilometres from Crikvenica, was in Šumadija. Klasnić of the Banovina was in Šumadija too. Krug was in Lower Austria. And Grad and Tišina, both Prekmurje parishes, were in Dalmatia. Three things were wrong underneath them. The region box existed only for the two Udine providers and now covers all seven. «Sv.» was being stripped and never expanded — GeoNames writes the patron out in full, so Sveti Jakov Šiljevica and Sv. Jakov-Šiljevica never met. And this builder had never read data/gazetteer.json at all, which is where a coordinate checked by hand lives with its reason beside it, so a hand fix could not reach the research map however carefully it was written. The audit now reports zero. It is [scripts/place-audit.py], it changes nothing itself, and it should be run whenever a place is added.

· Corrected in place: /map/

was The reconciliation published fourteen anchored disagreements this morning, the first of them Antonius Defranceschi — tree 1814–1893, archive 1893–1893.

now They are two people. Antonius of Gologorica died in 1893 aged seventy-nine. The other Antonius was buried at Pula as an infant in 1893 — born and buried in the same year, which is why his row reads 1893–1893. Twenty of this archive's burial rows look like that, because a burial register of a child gives both dates at once. The anchor test was written to stop exactly this: two people of one name and no shared year are two people. It did not occur to it that two people can share one year and still be an infant and an old man. The test now refuses any pair where one life is five years or less and the other thirty or more. Three rows went with it — Antonius, a Giovanni aged three against a Giovanni of seventy, and a Maria who died in infancy against a Maria who lived to forty-seven. Fourteen becomes twelve, and the twelve are then sorted by what kind of disagreement they are — seven that are a year of age-arithmetic, two that are too large for it, and three where the death year itself is disputed. Only the last kind needs a register. Found while working the fourteen one by one, which is what the work-list row asked for.

· Corrected in place: /live-tree/

was The reconciliation published 668 people only in the tree, offered as «where the next month of work is».

now Twice wrong, and the second time on this very page. First: 486 of them are inside David's own placeholder pens — twenty-three country folders with town folders beneath, under a node he labelled «for Investigation (not real)». He had already tested and parked them. Counting that as untouched work was backwards. Then the remainder was announced as 113, and that was wrong too. The join compared the whole set of given names, and the live tree stacks every variant it has ever seen into one string: «Johannes Giovanni Gian Battista de Franceschi» against this archive's «Giovanni Defranceschi» is four names against one, so the top of David's own male line was reported as a stranger — and this page said of him, in as many words, «that is not a namesake to be tested, it is a hole». He is in the roster and in the spine, born 1750 at Gologorica and dead 1829, exactly as the tree has him. The test is now: the same set of given names, or one name in common and a year in common. It is in the builder rather than in prose, so the page cannot drift from it again. The honest figure is 84. They stand at Ližnjan, Crikvenica, Gračišće, Gologorica, Brooklyn, and two beside Pfullingen.

· Corrected in place: /live-tree/

was That the forty burials with no parish «cannot be placed from anything in this repository», because the Istrian sweep's per-record output — ark by ark — was never written to disk, and that re-harvesting forty personas from FamilySearch was «half a day».

now data/parish-by-record-2026-09.txt has been in the repository since 8 September: ninety-three parishes, and under each of them the 952 persona ids the sweep read there. Only the ninety-three-row summary was ever published, so the file went quiet, and a row was then written asserting it did not exist. All thirty-six placed in an afternoon with no search at all — twenty-five on the id the sweep already held, eleven on its twin. The lesson is not about FamilySearch. A file that nothing links to is a file the archive will shortly claim it never had.

· Corrected in place: /worklist/

was Person pages carried, in as many words, «This person has no recorded year», «born about 1749 — estimated, not recorded», and «No family is drawn for this name, and no record stands behind it… the name reached this project through a chart, a narrative source or the family tree».

now For ninety-two of them that was false, and the evidence was in this repository the whole time. Nine of Valentino's children. On 6 September the Liber Baptizatorum of Fiume was read and the fourteen children of Valentino Defranceschi were written to site/src/data/valentino.json. No build script and no page ever read that file — for eleven days. Five of the children had pages estimating a birth «about 1749» over the top of an exact baptism date sitting on disk; four had no page at all, and their baptisms exist in no index anywhere, so this archive was the only place they survived. The file's fault was that it had no key to a person: it writes «Josepha Catharina» where the archive has «Josepham Catharinam Defranceschi», the accusative the clerk used, and no name match was ever going to join those. And eighty-three more, for a different reason. The «no record stands behind it» banner tested only whether a person had a household chart or a FamilySearch index strip. It never looked at their records. So Cesare De Franceschi was told no record stood behind his name two sections above his own headstone — «Headstone – 30 Jan 2005 · Gungahlin Cemetery and Crematorium» — and eighty-two others like him. Every gate was green while this was true: links, anchors, living people, the work list, the errand ledger. All of them look at what IS on the page. None looked from the evidence back to the person. One does now, and it refuses the build.

how scripts/check-evidence.py, wired into npm run build as check:evidence. It reads the rendered site, not the data, because the failure it guards against happens between the two. Verified by reintroducing both bugs: the first fails 9, the second 88, each exits 1.

· Corrected in place: /who/

was «Every name below was reconciled by hand from the forms that actually appear in these registers — not from a dictionary.»

now 130 of the 415 forms on the page — thirty-one per cent — appeared in no record this archive holds. The Croatian column was the worst of them: 31 of its 48 forms had never been seen written down here. And dozens sat under the wrong language outright. Antal and Ilona are Hungarian and were filed as Italian. So were Božo, Tomislav, Jerolim, Pasko, Terezija, Karlo and a dozen more Croatian forms, and Steven, Jasper, Noel, Dominic and Benedict, which are English. Bartholomaeus was filed as German and is Latin; Srećko as Italian and is Croatian. And there was no Hungarian column at all — in an archive whose own place-strings read «Vitipolis, Modrus-Fiume, Magyarország», for a family whose Fiume parish was in the Kingdom of Hungary and whose registers went over to Hungarian in 1895. Sixteen Hungarian forms were already sitting in the records unrecognised. Deleting the dictionary forms would have been the wrong fix. A form nobody here has met is still the right thing to type into a Hungarian or a German index tomorrow. So the table now keeps both kinds and says which is which — 419 of 645 cells attested in a record, 226 marked as lookup aids — across seven columns and 73 names, twelve of them new (Giorgio, Giulio, Elisabetta, Angela, Sebastiano, Bernardo, Leonardo, Pasquale, Mario, Marta, Mattea, Oliva), every one of which occurs in these records and had no row. Two wrong equivalences went with it: Marta is Martha, not a form of Margarita, and it has its own row now; Agneza is Agnes, not Agata, and it is gone, because this archive has never met it.

how scripts/name-forms-audit.py, wired into npm run build as check:names. It looks every form up in the roster, the register, the households, the burials, the headstones and the four FamilySearch harvests, and refuses the build if a form stored as attested is in no record. Verified by marking Gergely and Gregor attested: exits 1 and names both.

· Corrected in place: /your-name/

was «Somebody should photograph that lintel. It has been on this archive's list of things to look for since the beginning, and it has just gone from a guess to a target.»

now The family photographed it in the field, and the pictures have been on the gallery ever since. The arms, and the Madonna beneath them — the oval cartouche in high relief with its mantling and plumed helm, on the façade at Gologorica. The builder's tablet, on the front of the house. The blazon, cut in stone on the tomb. The same arms, worn almost away, on an older stone in the same ground. And the house itself from the lane. The page that argues the carving exists was the one page not showing it. They are on it now. What is still genuinely wanted is narrower and worth stating exactly, because the retraction should not swallow the real gap: every frame the archive holds was taken in flat light from wherever a person could stand, and none of them settles the fruit — which is the whole argument, Carlo's pomegranate against the cousins' turnip. The shot list has been asking all along for the cartouche straight on, filling the frame, in raking light.

how Found by David, reading the page. The same fault as the valentino.json one earlier the same day and the Trove key one before it: the archive holding a thing and the page saying it does not.

· Corrected in place: /arms/

was That the given-name table was the instrument behind the joins: «The normaliser behind the household pages now carries 73 canonical names across 645 recorded forms, in Latin, Italian, Venetian, Croatian, German and English.»

now Nothing read it. givennames.json was imported by two pages that display it and by nothing else — not by build-dossiers.py, not by lib/people.js, not by any script that joins a record to a person. Both normalisers in this archive fold the surname and stop: · _nm() lowercases, strips diacritics and maps defranceschi → de franceschi. · norm() in lib/people.js does the same. Neither touches a given name. So Josephus, Giuseppe and Josip were three strangers to every join on this site, and the page reconciling them said it was already fixed. 327 of the 1,122 person pages are in a group of this kind — 98 groups. Ivan Defranceski, Ivan de Franceschi, Joannem Defranceschi, Joannes Defranceschi, Giovanni Defranceschi, Giovanni de Franceschi and Zuane De Franceschi are seven pages. The fix is a cross-reference, not a merge. Identity here is by name on purpose, and collapsing Ivan into Giovanni by machine is the join this archive spends its time refusing. Each page now names the others and says plainly that none of them is claimed to be the same person.

how Found by David asking whether the archive actually searches with these forms. It does not do that either — see the work list.

· Corrected in place: /name/ · /your-name/

was That the first search of The National Archives at Kew found «nothing of this family», outcome «nothing». And, at the top of the front page, that this is «A family written down under eight sovereigns».

now This site has never been about one family and the site says so itself, two screens below that sentence: «There is no single Defranceschi family. Six families carry the name and this archive follows all six.» The front page contradicted its own card, and a Kew search written by that wrong measure threw away everything it found. Seven of those records are the name in Britain and belong here: a Franceschi in the State Papers for Venice in 1677; Mark Franceschi, master of the Amabile, taken as prize and appealed at Minorca in 1805; three Great War medal cards for Lorenzo, Ernest and William C. Franceschi; and Lorenzo Franceschi of London, from Italy, naturalised 4 April 1934, wife Agnes Elizabeth. Eleven more are Franceschin- — the diminutive, a different surname — and are filed as adjacent rather than discarded. Britain had no entry on where the name went despite a Cardiff marriage of 1876, a Cardiff recital of 1896 and a barque out of the Bute Docks in 1877 all sitting in these pages. It has one now. The lesson is about scope, not about Kew. «Is this our bloodline?» is the wrong question to ask of a record while collecting; it is the right question only when claiming. Collect on the name, claim on the evidence — and this archive had the two the wrong way round for an afternoon.

how Caught by David, in one sentence: «this site was never for 1 family, its always been for the defranceski name and its different lines / origins.»

· Corrected in place: /searched/ · /

was Thirteen spellings were declared in scope, among them Franceschich, and the drift chart drew it as a branch of De Franceschi from 1850 with the note «Slavic -ić written the Austrian way» — that is, Franceschi plus a Slavic suffix.

now It is the reverse, and the name is somebody else's. Franceschich is the Italianised spelling of Frančić, the Croatian patronymic of Franjo — Francis's son — formed from the forename directly, exactly as Franceschini and the other diminutives are. It converges in spelling with this surname and shares nothing else. Twelve spellings are now in scope; six are adjacent. The row has been taken off the drift chart as well as the list, because a chart of how this surname branches cannot draw a branch that is not it. Nobody is displaced. No person in this archive carried the spelling — it was a claim about the language, not about a household.

how Found by the Crikvenica marriage film, which is what a complete reading is for. Two witnesses spelled Franić / Frančić appear in the 1888 marriages, at entries 6 and 18. Under the old list they were either two hits or an error on the list, and it could not be both. David ruled on 20 September 2026: «frančić is adjacent, take franceschich off the list» — so the seven-year negative for 1883–1889 stands unqualified, and the list loses a spelling it should never have had.

· Corrected in place: /name/#scope

was A search row of 20 September announced Svetvinčenat as a discovery: «It is filmed, it is filmed from 1568, and nobody here has opened it», called it «a sixth Istrian parish now reachable» and then «the name in a seventh Istrian parish», and claimed that «the oldest register this archive can open is no longer Gračišće's».

now Every part of that was wrong, and the archive's own files said so. The parish was not unopened. This site carries a section called «Reading the Svetvinčenat register, page by page» — nineteen openings of DGS 005497884 read band by band, covering 1666–1707, with twenty occurrences of the surname that no index holds. It carries the 1628 baptism of Elena, daughter of Francescho di Franceschi, which moved the whole Istrian sequence. The roster has seventy-five people at Svetvinčenat; the atlas place record says nfilms: 9; the families page says «at least twelve distinct fathers of this surname in one small town between 1657 and 1737». It was never a «sixth» or «seventh» parish, and the count contradicted itself in consecutive rows. And the archive has never called Gračišće's 1644 its oldest openable book — that sentence was invented to make the find look larger. How it happened. The work-list item said «This archive has six lines and none of them is San Vincenti» — which is true of the lines, and says nothing about the records. That was read as «the parish is untouched» and nothing was checked against the roster, the atlas or the searched log before publishing. It is the same failure this archive caught in the Gračišće work-list note the same afternoon — a claim about holdings made without opening the holdings. What actually survives of the work. The reading itself stands and is worth keeping: the 1709–1764 and 1764–1791 books are genuinely the next two on that shelf and had not been read, so Gio Batta, Giacomo, Gasparo, Battistina, Dominicus and Antonius are real additions after 1707, where the September sweep stopped. The rows have been rewritten to say that and only that.

· Corrected in place: /searched/

was The Fiume household page was headed «Valentino's fifteen children» and counted ten sons and five daughters. Work-list item 142 built on it: the children «fall in 1737, 1739, 1741, 1742×3, 1745, 1751…», so the holes were «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 same page dated the family's rise to «by the 1750s it calls the father D.nus Valentinus and stands Illustrissimi over the font».

now There are sixteen children, and the silence is not a silence. Image 146 of film 005497925, under «Die 13 8bris 1747»: «Franciscam Matthæam filiam legitimam D.ni Valentini, et Helenæ jugalium Defranceschi, natam 10 hujus» — a daughter born 10 October 1747, inside the supposed five-year gap. She is in no index anywhere, and she is not Matheus Franciscus, who was baptised 17 September 1739 and carries his own ark. Why it matters more than one name. The gap was the premise of the archive's most promising open question: if the household went quiet for five years, an unrecorded son could be hiding there, and that son could be the Giuseppe who heads the Rijeka chain. The quiet was an artefact of the index plus an incomplete read — and the first stretch of the gap ever swept frame by frame produced a child on the ninth frame. The hypothesis is not dead; its foundation has simply been replaced by evidence. And the styling is three years early. The same entry already reads D.ni Valentini, with Ill.mus and Nob. godparents, in 1747 — not «by the 1750s». What survives. The family's own tradition of ten sons is untouched: the sixteenth child is a daughter. The household now stands at ten sons and six daughters.

· Corrected in place: /fiume-1626/#valentino

was The Rijeka household Giuseppe Defranceschi × Francesca was published with four children, 1772–1775, built «from index entries only — this archive has not seen the page», and listed both «Valentino Defranceschi, b. 1772, bur. 22 Aug 1775» and «Valentinus Franciscus Xaverius Defranceschi, b. 22 Nov 1772» as separate children.

now Five children, not four, and one of the four was the same boy twice. All of this household's baptisms were read on film 005497925 on 20 September 2026. The duplicate. Valentino and Valentinus Franciscus Xaverius are one child — baptised 23 November 1772, «heri natum», and buried 22 August 1775, aged two. The register settles it rather than the arithmetic doing so: the next son, baptised 12 March 1776, is christened Valentinus Josephus Gregorius — the name given again after the first died, which is what these families did. The two that were missing. Valentinus Josephus Gregorius, 12 March 1776 and Hellena Ludovica, 8 February 1778 were in no household on this site. Both were read on the page today, with their godparents. And the note is retired. «No register has been read» was true when it was written and is the reason the household was trusted only as far as it went. It has been read now, and the household says so instead. The counting error was invisible from the index, because the index holds a baptism and a burial under two different latinisations of one name and has no reason to join them.

· Corrected in place: /households/

was The Atlas carried «Pagadian» at 7.826 N, 123.437 E — Mindanao, in the Philippines — for a record whose own label reads «Pag, Zara, Dalmatien, Österreich». It carried «Basel» at 47.558 N, 7.573 E — Switzerland — for Pietro De Franceschi. And it carried «San Lorenzo» at −25.340, −57.509 — a suburb of Asunción in Paraguay — for two people recorded at San Lorenzo, Santa Fe, Argentina.

now Pag is in Dalmatia, Bale is in Istria, and San Lorenzo is in Santa Fe.

· Corrected in place: /atlas/

was After three mis-placed pins were corrected earlier the same day, the Atlas still carried five more. «Kabwe» in Zambia held Giacomo, John and Robert Defranceschi, whose record reads «Broken Hill, New South Wales, Australia». «Mitchell» stood in South Dakota for a record that says «Mitchell, Canberra, ACT»; «Parkville» in Maryland for «Parkville, Victoria»; «Athens» in Greece for «Athens, Clarke County, Georgia»; and «Dolo» in Ethiopia for «Dolo, Venezia, Veneto».

now All five now sit where their own records say, and the worst of them was out by eighteen thousand kilometres.

· Corrected in place: /atlas/

was Work-list item 143 was published with a bounded, ready-to-run instruction: «The Pazin volume is enumerated and anchored, ready to run… Index 134 = January 1741. Index 142 = 1746. Index 160 = October 1747 – January 1748. Read frames 128–165.» Those years were read off the pages, not interpolated.

now They are 1711, 1716 and 1717, and the window is frames ≈194–204. In this clerk's hand the 4 has a nearly closed bowl and a flat crossbar and is not distinguishable from his 1 at page scale. Each frame read as a plausible sentence on its own; only the run was impossible, and nobody checked the run. Frame 143 broke it — its right-hand page writes 1717 twice, cleanly. A frame eight earlier cannot be 1746. Re-anchored by reading: 178 = 1733, 198 = 1742, 205 = a microfilm resolution target, which is a volume boundary. About 2.2 frames to the year. Why this is published rather than quietly edited. The wrong window was not a note to self; it was written as ready to run, with arks on disk, for whoever picked the item up next — and it would have cost them thirty-eight frames of 1711–1717 before they noticed. An instruction that confident deserves a correction that loud. And it sharpens this archive's own rule. The Method page already says anchor by reading, not by interpolating. That was obeyed here and was not enough. Reading is not enough when the digit is ambiguous: a year is anchored only when it is written unambiguously, or cross-checked against a neighbouring frame, and the sequence itself is the cheapest check of all — 1746 at frame 142 beside 1717 at frame 143 is not a close call, it is impossible, and it was sitting in the notes. Nothing was read on the strength of the wrong window. The error surfaced while setting up to run it.

· Corrected in place: /worklist/

was The work list described the Svetvinčenat births index as an alphabetical table to be swept letter by letter, and listed what remained as «A, I, most of M, N, O, P, R, S, T, V, Z» — «which is where Antonio, Domenico, Lucia, Maria, Nicolò, Pietro and Zuanne would be».

now The index runs B to P and no further, at either end. Read end to end on 21 September: it occupies images 196–211. Image 195 is register text, so there is no A page. Image 212 is a loose sealed certificate and image 213 is the film's «END OF ITEM» card, so there is no R, S, T, V or Z. The letters that exist are B, C, D, E, F, G, L, M, N, O, P. The missing letters are the ones this family is named in. Antonio and Zuanne are the two forenames Svetvinčenat is known for here — «M.ro Ant.o di Franceschi di Piero» of 1667, and the Antonio Defranceschi who led the castle households to their exemption of 1780. Neither can be found through this index at all. Why it matters more than a missing page. The list said those letters were unread. Had anyone swept them, found nothing and written «no Antonio in the index», the archive would have recorded an absence that was never a presence to begin with. A finding-aid that stops silently produces false negatives, and they are indistinguishable from real ones. The rule this yields: establish a finding-aid's extent before trusting any silence in it. Also corrected from the film's own title card: the volume is «Matična knjiga rodjenih 1703–1763», not the catalogue's 1709–1764.

· Corrected in place: /searched/

was Published this morning, from the Svetvinčenat index alone: «The families page said «1733 Maria, of Giacomo and Anzoletta»; carta 66 said «Batta de Fran.chi et di Anzoletta» in 1728 — two husbands for one forename… Batta and Giacomo are two different men, and Anzoletta was carried from one entry to its neighbour.»

now Carta 83 was opened this afternoon and it says Anzoletta in its own hand. «Maria fig.a di Giacomo Defranceschi di Pietro, e di Anzoletta», 4 November 1733. There is no neighbouring entry to have carried it from: it is the first entry on the leaf. The half that was right is that Batta and Giacomo are two different men — carta 66 carries both their daughters, one day apart. The half that was wrong was the explanation. The reasoning ran: the same wife's name appears against two husbands, therefore one of them is a copying slip. That is a perfectly good hypothesis and it was published as a finding. What the leaves actually leave standing is harder. Giacomo's wife is Elena on carta 66 in 1728 (read at magnification), Anzoletta on carta 83 in 1733, and this archive already carried a Maria as his first wife and an Anzola in 1737. Three or four names in ten years is not one man's marriages — it is most likely two Giacomos, which is exactly what the 1733 clerk was signalling by writing «di Pietro» after the name for the first time. The rule this cost. An index entry is a summary somebody else wrote, and a disagreement between two summaries is not evidence about the people — it is evidence about the summaries. The conflict was real; the resolution had to come off the leaf, and the leaf was four clicks away.

· Corrected in place: /searched/

was «Casparus filius legitimus Jacobi Defranceschi, et Francisca ejus legitimæ uxoris» — 1 June 1743, film 005497885 image 134, read on the plate 23 September 2026. Godfather given as «D.nus Dominicus Xaverii Bosponi».

now The word is Pascasia. Re-read at 8× on 28 September and set beside the same clerk's own hand, twice. · Against his «Franciscus». Carta 119v of the same opening carries «Franciscus filius legitimus Caroli Varnier», 8 October 1743. The clerk's F is a crossed single loop followed by narrow minims — r-a-n-c-i-s-c. The word on carta 119 opens with a doubled bowl closing into a descender and runs a-s-c into a broad round vowel. «Franciscus» has no round vowel in its middle at all. The two words share only their last two letters, which is how the misreading happened. · Against the register's own «Pascasia». Image 143, 31 March 1746: «Martina Antonia filia legitima Jacobi Defranceschi, et Pascasia ejus legitima uxoris» — dark ink, a full parental entry, and a name this archive already reads correctly. Set beside image 134 at the same magnification, the two words are the same word, stroke for stroke. So Casparus is Pascasia's son, and [h107] Jacobo × Pascasia begins in 1743 — not 1746, and not the 1750 this archive first had. What it does to the two-men case: it takes 1743 away from it. The page argued that «Francisca Marcario brackets Pascasia by five months on one side and a week on the other» and that «no widower remarries and is widowed again inside eight days». With 1 June reading Pascasia there is one wife in 1743, not two — Pascasia in a full baptism on 1 June and as godmother on 24 November — and then a single «Mar-/:cario uxor Jacobi Defranceschi» on 1 December. One godparent line, standing alone. That is precisely the evidence this archive has twice convicted: «Pascasia uxor Caspari Defranceschi», 16 December 1751 and 3 February 1752, both provably wrong about the husband. The two-men case is not withdrawn. It is moved to where it was always strongest and where nothing has been re-read: 1752–53 — «Francisca uxor Jacobi Defranceschi» on 6 February 1753 against «Jacobus Defranceschi, et Pascasia ejus uxor» in one joint phrase on 29 April 1753, ten weeks apart, with Pascasia bearing his children in 1746, 1747 and 1750. (inf, and unchanged in strength — but it no longer draws on 1743, and the page said 1743 was what settled it.) A second misreading in the same entry, caught the same way. The godfather is not «Dominicus Xaverii Bosponi». He is «D.no Francisco Xaverio Profumo» — the chancellor of the parish — named in full on carta 119v of the same opening as «D.ni Francisci Xaverij Profumi Cancelarij», whose wife Veneria stands at another font on that very page. Three of those four words were wrong, and the right ones were legible one leaf away. And «Bosponi» was not one slip in one entry. It was the man's name throughout this archive. Checked on three frames of the same film, in three different years: carta 119 (15 June 1743, «Francisco Xaverio Profumo»), carta 119v (8 October 1743, «D.ni Francisci Xaverij Profumi Cancelarij») and image 147 (27 October 1747, «D.ni Francisci Xaverij Profumi»). All three read P-r-o-f-u-m-, with the tall f and the -um- tail. None of them has a B. Corrected in four files — the baptism of Antonia di Jacobo Defranceschi (1747), Faustino di Joannes Baptista Defranceschi, the households file and the searched file — and with it the name of a man who is not a Defranceschi at all, but who stands at this family's fonts across at least five years and whose son does the same. A one-name study that misnames the neighbours is misnaming its own evidence. One more, found in passing and verified the same way: the chaplain published as «Martino Zolonta» is «Martino Zolonda» — the letter is a d with a full ascender loop, read at 5× on image 143 where the ink is dark.

how Both misreadings are the same mistake, and it is not bad eyesight. «Francisca» was already in mind — she is Matteo's wife, attested on this film in 1741, 1742 and 1752. «Pascasia» was not yet known to the archive in 1743; it first meets her in 1746. The clerk wrote an unfamiliar name and the reader supplied a familiar one. The register was never ambiguous. It was read against what was expected. What caught it was refusing to read a word in isolation. Neither the 8× magnification nor the transcription settled anything on its own. What settled it was finding the same clerk writing the same two candidate names elsewhere in the same volume and putting the three side by side at one magnification — his «Franciscus» on the facing leaf, his «Pascasia» nine images on. A hand is only legible against itself. The same session withdrew two of its own readings before publishing them by exactly this test — «Bascovicia» on image 135, twice — which is the method working in the direction it is supposed to. The scope was found by asking the obvious next question and not liking the answer. Having corrected one godfather, the cheap check was «how many times does this archive say Bosponi?» — four records across four files, all resting on the same unverified reading, none of them ever re-read. A misreading is rarely confined to the entry where it is caught, because the reader who made it went on reading.

· Corrected in place: Casparus di Jacobo Defranceschi (1743) and Jacobo de Franceschi of Svetvinčenat — roster.json

was «It carries the two Bakar newspapers, which is the sharpest reason to work it, and both are empty: «Galeb» and «Kvarner» return zero in every spelling across two complete runs, with the controls passing. A genuine absence, not a scanning failure.»

now «Galeb» is not empty. It carries the name once, and the entry is a good one. Galeb (Bakar), 15 July 1879, image 2: «U našoj posestrimi Istri izabran je dr. D. Vitezić, dočim je drugi narodni kandidat dr. Klodić u seoskih obćina podlegao Talijanu Defranceskiju» — the Croatian national candidate was beaten in the rural communes of Istria by the Italian Defranceski. (doc.) And the portal finds it without difficulty. Searching «Defranceski» with the title restricted to «Galeb (Bakar)» returns exactly one result — the same entry. The per-title search works; the claim was simply wrong. Kvarner does return zero, tested the same way on the same day. So half the claim stands and half does not, and the half that does not is the half that mattered — Bakar is this archive's own ground, and a «genuine absence» there was being used as a finding. The likeliest cause is the spelling. *The all-papers counts in that note are «defranceschi» 46 and «defranceski» 4 — and the Galeb hit exists only under the -ski spelling. If the Galeb run tried «franceschi» and «defranceschi» but not «defranceski», it would have come back empty and honest. But the note said «in every spelling», and that is the sentence that was wrong.

how A negative finding is a claim, and it needs the same discipline as a positive one. «A genuine absence, not a scanning failure» is a strong statement — it tells the next reader to stop looking. It was published on the strength of two runs whose spelling list is not recorded, which means nobody could check it without redoing it, and for weeks nobody did. What would have caught it: writing down what was searched, not just what was found. This archive's /searched/ log does exactly that for most sources — source, terms, date, outcome. The Galeb and Kvarner runs went into a prose note instead, where «every spelling» could stand in for a list nobody kept. The rule this earns: a nil result names the exact terms tried. «Nothing under X, Y and Z» is a finding. «Nothing in every spelling» is an assertion about a list that no longer exists.

· Corrected in place: Sources — sources.json, the dnc-nsk entry

was «What it means, if «q.m» is right: Antonio was dead by November 1780.» — read from frame 86, where an ink blot sits on the decisive glyph between «Gregorio Defrance—» and «Ant.o». This archive compared the mark to the undoubted «q.m Vitt.o» on the next line, judged it the same, and published it as doc for the name with the blot stated.

now He was alive two and a half years later, and he had a title. Film 005497886, frame 93 — the same volume, seven frames on — gives him twice in six weeks, in a hand that is not blotted: · 13 March 1783: «Vittorio figlio di Capo Antonio de Franceschi contrasse li primi sponsali con Eufemia figlia di Mattio Bergamo.» · 28 April 1783: «…qual esposero alla mia presenza di Capo Antonio de Franceschi, di Mattio Percich q. Piero…» «Capo», not «quondam». And the same clerk marks his dead in the same sentences — «Fosca figlia del deff.o Antonio Gigante», «Mattio Percich q. Piero», «Capo Mattio Buzzalovich q. Zuanne». The living title and the death-marks sit side by side in one line, which is as clean a control as a register ever offers. So the blot reading is withdrawn. The name Gregorio Defranceschi on frame 86 stands — that part is legible. What he was called after it does not, and this archive no longer says that Antonio was dead in 1780, nor that Gregorio's father was. What went wrong. The comparison was made against the right line and still reached the wrong answer, because the glyph it was compared with was never fully visible — the blot covers it. A shape that "looks like" the control is not the control. Where ink covers the decisive letter, the honest reading is no reading, and that is the rule this archive takes from it: a hand is only legible against itself — and not at all where the ink has gone.

· Corrected in place: Gregorio Defranceschi, Antonio de Franceschi, the Svetvinčenat line, Searched — all published the same morning

was «Moncalvo is in Piedmont. Two hundred miles from Istria, written on four separate documents by different clerks.» — used as the decisive disproof that Arolsen's «DEFRANCESCO, Antonio, born 31.1.1909 zu Moncalvo, died Dachau» could be the Anton Defrančeški born 1909 at Gologorica whom the Belgrade victims' census records.

now Moncalvo is the Italian name of Gologorica. This archive's own Gologorica page states it plainly: «Gologorica is the bare hill in Croatian; Moncalvo is the same thing in Italian. Which of them a man wrote told you where he stood.» The word appears in thirty-three of this archive's files. So the German clerk who wrote «geboren 31.1.09 zu Moncalvo» was writing the Italian name of the very village the Belgrade census names. There is also a real Moncalvo in Piedmont — which is what this archive reached for, without checking a page it had written itself. ⚠ And the two details flagged as puzzling now resolve, both at once. The Arolsen man's committing authority was Sipo Trieste, and the Dachau entry register gave his nationality as «ITALIAN, YUGOSLAVIAN». Those were called «Adriatic markers, not Piedmontese ones», recorded as an oddity and left. They are not an oddity. They are what an Istrian from Gologorica looks like in a German camp register. What this does and does not establish. It withdraws the disproof. It does not turn the identification into a fact:

· Corrected in place: Anton Defrančeški's page, Searched and item 160, all written earlier the same day

was «Anna, legitimate, daughter of Josephus Kalanj and Rosalia née Perković, at Rijenovica house 22» — carried in fifteen files, including the pedigree's own wall note: «Rosalia née Perković of Rijenovica house 22 is, for now, in no archive at all.»

now David asked whether Kalanj deserved a page of its own. Building it found this. There is no Rijenovica. The word returns zero hits in this archive's Croatian gazetteer, while Klenovica is in it and stands in thirty-four of these files. The register, read again at 6× from the plate this archive already held: entry 2 of 1864, locus domicilii — «Kljenovica N° 22». Kljenovica is the older spelling of Klenovica. (doc.) ⚠ And that changes a fact about the direct line. This archive's Crikvenica page already says, of a photograph of Klenovica: «House 22 is somewhere in this line of roofs — Anton Rudolf Defranceski was born in it on 20 January 1890.» His mother Anna was born in the same house, twenty-six years earlier. The archive has held «house 22» at both ends for months and never joined them, because it had misread the name at one end. ⚠ A second thing was dropped from the same entry. The parents' column reads «Josephus Kalanj et Rosalia nata Perković ejus uxor, Brinjensis Parochiæ ruricolæ» — country people of the parish of BRINJE. No file of this archive carried that, and it is the single most important phrase in the entry: it says where the Kalanj came from before the coast. How it was caught. Not by reading harder — by comparing two archives. The Blažević Archive's Kalanj page reads the same entry as «Klenovica N° 22» and names Brinje as «the single largest hole in this family». Two family archives held one baptism and disagreed about the village, and nobody noticed, because each read only its own page. That is the cost of a boundary drawn to prevent duplication.

· Corrected in place: The ancestors' walls, the Crikvenica house, the plate caption and twelve more files

was «Of a family with a palace on the square facing the bishop's; by 1848 a Salomon sat in the parliament at Vienna.» — carried unsourced, and repeated on the maternal-line page built this morning.

now David downloaded the list. There is no Salomon in it, and a Defranceschi is. The official roll of the first Austrian Reichstag, 1848–49 — 422 deputies with name, character, electoral district and land — reads, verbatim: > Defranceschi · Karl · k.k. Rathsprotokollist · Pisino · Küstenland ⚠ And the negative is properly made. 437 lines, 65 surnames beginning with S, and zero matches for Salomon or Salamon in any spelling — with the control passing, since the S-names parse correctly and are listed. No Salomon sat in that parliament. What the family remembered was true; it was attached to the wrong surname. There was a seat in the Vienna parliament of 1848 in this story — it was their own. Carlo De Franceschi of Gologorica, the historian, sat for the district of Pisino — Pazin. ⚠ The archive already half-knew it and had never had the constituency. *Carlo's page has carried since 13 September that he «sat in Vienna» in 1848 and was a «Rathsprotokollist des Landrechts zu Rovigno» — but that came from a newspaper attacking him, the Journal des Österreichischen Lloyd of 7 September 1848. This is the official roll, it confirms the office, and it adds the thing the newspaper never gave: the seat he was elected for. So the Salomon note loses its parliament and keeps its palace — the Salamon palace at Gračišće is real, sixteenth century, dated 1570 — and the parliament goes where it belongs.

· Corrected in place: The Salomon maternal line and the pedigree note it came from

was «M01069560 – M01069944 — A twentieth-century Status Animarum of the parish of Gologorica, on printed forms, arranged by hamlet — Riemunjica, Želeski, Sokolići — with a repertory indexed by age bands from 1890 to 1980, and an Italian-period stamp reading Comune di Fianona. It carries deaths into the 1970s.» — and, in the same block, that walking film 215 «gives three books, not one».

now That range is three separate books, and the film says so on its own title cards.

· Corrected in place: The direct line, the walk of microfilm ZM34K-215

was «Antonius Defranceschi, b. 14 April 1814 — Son of Leonardus (house 6); married Francisca and became head of the house.» — and the pedigree drew him as the only child of Leonardus × Matthea.

now The 1826 status animarum of Gologorica names his parents, and they are not Leonardus and Matthea. Three independent statements, all on image M01069328, all in the same hand: 1. The name column — «Antonius Defranceschi filius Usualdi et Apoloniæ». 2. The adnotatio — «hunc in filium adoptivum accepit patruus Leonardus, et locatus hic sub N.º 6» — his father's brother Leonardus took him as an adopted son and settled him here under number 6. Patruus is specifically the paternal uncle, which is what makes Usualdus a brother of Leonardus rather than an unrelated man. 3. The foot of the same page — «Usualdus … pater Antonii», written under a house marked «non habitata». His own house at Gologorica stood empty when the census was drawn. So the line runs sideways where this archive drew it straight. Leonardus and Matthea are now recorded with no children; Antonius moves to a new household under Usualdus Defranceschi and Apollonia, both of whom are new to this archive. Every descendant already published under Antonius keeps its place — he did head house 6 — but his ancestry above that point changes. That Leonardus and Matthea were childless is the obvious reason for the adoption and is not stated by the register. It is recorded as inference. (doc for the parentage and the adoption; inf for the childlessness.)

· Corrected in place: The Gologorica line — the pedigree, and every descendant drawn under it

was «Aloysius Defranceschi, b. 8 May 1832 — son of Josephus & Lambertina (house 7); recorded as «natus Polonia» (born in Poland)», with place: Poland on the person and the line south on the roster. This archive repeated it in three files and drew a conclusion from it: that the household had moved.

now The word is VOLOSCA. The 1826 status animarum of Gologorica, house 7, reads «Aloysius … filius natus Volosca» — read at five times magnification, where it is not ambiguous at all. And it is the obvious answer, once seen. Two lines above him on the same page his mother is entered «Lambertina uxor nata Peschle» with «Ex Volosca» in the provenance column. She went home to her own town to be delivered — which is what women did, and which explains a birth away from the parish without needing Poland at all. «Polonia» and «Volosca» are not similar words. The error was almost certainly a reading of the initial V as a P, made once and then repeated in three files without anyone going back to the page. ⚠ One thing is NOT corrected. The register gives his birth as 4 May 1832 where this archive carries 8 May. A single page-scale reading of a digit does not move a date, and his baptism is in MKR 1784–1880 on the same reel. Left open. And this is the second time in one day that Volosca has turned out to matter — David asked this morning whether its books had ever been opened. They have not, though the archive has held five waypoints for them all along. (doc.)

· Corrected in place: The Gologorica line, and his roster row

was Six children were filed under «Joannes Baptista Defranceschi × Mariana», and each carried the note «Daughter/Son of Joannes Baptista (casa 36) and Mariana».

now The baptism register names the mother three times, and she is not Mariana.

· Corrected in place: The Gologorica line — the household of Joannes Baptista at casa 36

was Two differences were logged as unresolved on 30 September and 1 October: the census gave Carolus as 17 October on one page and 19 October on another, against this archive's 17 October; and Franciscus Xaverius as 25 February 1816 against this archive's 19 February. Nothing was changed on either.

now The baptism register gives both dates for both men, and explains the difference. > «Die 19na 8bris 1809 · Carolus Josephus Defranceschi natus die 17ma hujus, vespere» — born the 17th in the evening, baptised the 19th. > «Die 25ta Februarii 1816 · Franciscus Xaverius Joseph Defranceschi natus die 19na hujus mensis» — born the 19th, baptised the 25th. Neither reading was ever wrong. This archive held the births; the census in places recorded the baptisms. Both are right, and the difference is the interval between the two events — two days for Carlo, six for his brother. ⚠ AND THAT IS A WARNING ABOUT THE WHOLE CENSUS. Its «Nativitatis» column cannot be assumed to hold birth dates. It gave Carlo's birth on page 14 and his baptism on page 89 — the same book, inconsistent with itself. Every date taken from that census is therefore a date of birth or of baptism, and which one is not knowable from the census alone. The four remaining date differences logged against the 1826 census on 1 October are most likely the same thing, and are left open until each baptism is read. (doc.)

· Corrected in place: The Gologorica line, and every date read out of the 1826 status animarum

was Three entries claimed walked: true over ranges of the 1784–1880 baptism book: M01069020–M01069036 (17 frames), M01069037–M01069051 (15 frames) and M01069052–M01069079 (28 frames) — sixty frames in all.

now Twenty-seven of those sixty frames were opened. The rest were not. The reading proceeded at roughly every third frame, which is a sample, not a walk. Why it matters more than it looks. walked is the one field in this archive's search register that tells the next person a book is finished. The [Record Atlas](../Record%20Atlas) states it plainly: «overstating it is the worst error available here, because it tells the next person a book is done when it is not». A sampled book marked walked is worse than an unread one, because nobody goes back to it. All three rows now carry walked: false and list the exact frames opened. The findings taken from those frames stand — Carlo's baptism, his brother's, the Jablich entries, the marginal death dates. What is withdrawn is the claim that the frames between them were read. The 1826 status animarum claims are NOT affected: those 119 pages were read one at a time and reported page by page, and that book is genuinely finished. Found by auditing my own walked flags against the frames I had actually cited. Nobody raised it.

· Corrected in place: The search register — three rows for the Gologorica baptism book

was Fifteen rows logged today for the Gologorica baptism book carry walked: true and the words «every entry read» — the flag that tells the next person a book is finished and that a nil can be trusted.

now The crop stopped at 62% of the page height, and these pages are written further down than that. It was caught at folio 124 (M01069138), where the 1863 entries are numbered: the crop showed ten and the page carries fifteen. Checking backwards showed the same fault on earlier folios, where it was invisible because the entries are not numbered:

· Corrected in place: The search log — fifteen rows · the Gologorica line

was The families page described the southern-Istrian family as «Southern Istria — Pula, Fažana, Ližnjan», resting on the name being «in the Pula registers from 1667».

now Those entries are from the Svetvinčenat register, DGS 005497884, and the parish carries the surname from 1657. Pula holds ten register entries of the name, only three accepted, and four of them are spelt Bellulovich Francessich — a different surname.

how The correction itself was made days ago and published on the lines chart. It never reached families.json, which went on asserting the original claim in its own title. This is the second time in this archive that a published correction has failed to reach the data behind it — the first cost forty-four people their parish. Found by auditing the chart's place labels against every place string in the register, the roster and the households, which is a check that should run on a schedule and does not.

· Corrected in place: /families/ · /lines/

was The lines page filed Corsica under «not ours».

now Nobody had checked. Not one search had ever been run in a French archive for this surname, in any commune. «Not ours» was a guess wearing the clothes of a finding — the same error as a false join, made in the opposite direction.

how Replaced with a stated position and its reasoning on a page of its own: a parallel, not a branch, expected to be unrelated and not yet tested. Dismissing a line without looking is exactly as unevidenced as claiming one.

· Corrected in place: /lines/

was A method note stated the persona endpoint caps out at about 1,100 records.