What the indexes actually hold

Holds: eleven forms of Saniger, when nineteen were needed

Test — Two records found within an hour of each other under spellings not on the list: SANIGOR (Hannah's baptism) and SANIGRE (her brother's marriage).

Every “not found” published here for this family was really “not found under the spellings I happened to think of”. The list now carries nineteen, and the failure was mine rather than any source's. Where it bites →

Holds: deeply, and it has been the most productive source of the fortnight

Test — Control on “Chatham”, May–December 1827: 4,733 articles. Date filters work once driven through the interface rather than the URL.

It produced Robert D'Arcy's seventeen death notices and his funeral, George Pitt D'Arcy's marriage and the only statement of his paternity, Joseph D'Arcy's obituary and his wife's name, and the Irish service of the 39th. None of that is in any parish register or census. Where it bites →

Holds: less than the record offices do

Test — FindMyPast, opened once. The 1851 Wakefield household, the 1831 marriage, and Berkeley baptisms 1792–1821 with parents named — all previously reported here as absent or untestable.

The commonest reason a record could not be found by this archive is that it was behind a subscription. Read every null below with that in front of it. Where it bites →

Holds: one partial index and no other

Test — HANNAH SMITH at Berkeley 1795–1815: nineteen results, and the only baptism collection returning anything for Berkeley is England Births & Baptisms 1538-1975, the FamilySearch index. FindMyPast's own Gloucestershire Baptisms covers Bristol parishes and does not reach Berkeley.

It is enough to put six Saniger generations on record. It is not enough to find Hannah, whose brothers of 1801, 1803 and 1808 are in it and who is not. Gloucestershire's parish registers are an Ancestry collection. Where it bites →

Holds: Roman Catholic parish registers, and nothing Anglican

Test — SMITH, marrying in Co. Wicklow 1813–1825: twenty-five results from exactly two collections — Ireland Roman Catholic Parish Marriages and Betham's Genealogical Abstracts. Seventy D'Arcy marriages across Dublin diocese 1814–1824, every one Roman Catholic.

An officer of a British line regiment married in the Church of Ireland, and those registers are the ones destroyed when the Public Record Office of Ireland burned in June 1922. This is the only holding on this page where the records are not merely unreached but gone. Where it bites →

Holds: richly — and its date filters cannot be driven from a URL

Test — A control on the word “Chatham” for May–December 1827 returns 4,733 articles.

Seventeen notices of Robert D'Arcy's death were found once the filters were set by hand. The archive is deep; the obstacle was an interface, and this is the only holding on this page where the limit was mine rather than the source's. Where it bites →

Holds: baptisms from the sixteenth century, with parents named

Test — One search. Jean Ward, 19 July 1754, father Joseph Ward, mother Martha Garden.

This archive reported South Leith as untranscribed and her baptism as unreachable, more than once and on more than one page. It was a statement about the free indexes and it was published as a statement about the parish. Where it bites →

Holds: to 1677 only

Test — The transcript's own end date.

One of the six Saniger generations in the tree can be tested here. The other five — 1679, 1711, 1744, 1769, 1805 — fall entirely outside it. Where it bites →

Holds: baptisms, 1780 and 1781

Test — Five common surnames across Hampshire 1778–1783: 46 Portsea entries, all baptisms — 20 in 1780, 25 in 1781, one in 1782, none in 1778, 1779 or 1783.

Joseph D'Arcy's baptism survives inside a twenty-two-month window. The absence of his brothers and sisters proves nothing whatever. Where it bites →

Holds: essentially nothing, 1800–1820

Test — William and Mary Smith across Gloucestershire 1800–1820: 289 entries from 75 parishes, and exactly ONE inside Bristol — a single entry at Stapleton.

St Paul, St Philip and Jacob, St James, Clifton, Bedminster, Temple, St Mary Redcliffe: absent at any size. Every Bristol search this archive has run was searching a city that is not in the index. Where it bites →

Holds: a fragment

Test — Three forenames of Smith across Somerset 1795–1815: 943 entries from over a hundred parishes, and TWO from Chew Magna, against a hundred at Walcot and sixty-nine at Bath.

The parish that now carries the best lead on Hannah Saniger's birthplace cannot be searched from a desk. Where it bites →

Holds: nothing

Test — Smith across Shropshire 1783–1792: 167 entries from the county — Shrewsbury 70, Hodnet 15, Ludlow 10 — and NONE at Oswestry.

Elizabeth Margaret Oliver's baptism, which the tree puts at Oswestry on 7 May 1787, cannot be tested here. Where it bites →

Holds: almost nothing, 1800–1820

Test — John Smith across the WHOLE COUNTY 1800–1820 returns TWO entries — one at Whitegate, one at Wrenbury.

The marriage of Samuel Sneyd and Elizabeth Oliver at Audlem on 22 October 1809 cannot be tested here. This is the emptiest county holding this archive has measured. Where it bites →

Holds: 1861 only

Test — John Smith born Staffordshire 1800–1805: 31 people, every one of them 1861.

Samuel Sneyd died in 1855 and his son had been in Australia since 1832. Neither was ever findable here, however hard anyone looked. Where it bites →

Holds: 1841 through 1901, and richly

Test — William Smith born Gloucestershire 1800–1812 returns entries from 1841, 1851, 1861, 1871, 1881 and 1891, including twelve inside Bristol for 1851 alone.

This is why the census, not the parish register, is what has actually moved this archive's Bristol work. The Saniger household at Oxford Road came from here. Where it bites →

Holds: complete enough to argue from

Test — Every Samuel Sneyd death registered in five years: exactly one. Every Elizabeth Sneyd in four: two, one of them at Stoke.

Strong enough that a single result across a five-year window can correct a family tree by a year. Used that way here. Where it bites →

Holds: unknown

Test — None possible. Each sits behind a bot check that this archive will not attempt to defeat.

Four sources that may hold answers and that have never been read. They are on the errand list for a person with a browser, not recorded as nulls. Where it bites →

The template that renders this table was a fallthrough chain — good, partial, absent, and everything else «cannot answer». So blocked, which means a source nobody has asked, because it wants an account this archive will not make, displayed as a source that had been asked and could not answer. Those are opposite statements: one is about this archive's choice and the other about the source's emptiness. One row also carried the freehand verdict «unusable as a control», which fell through the same way. The five words are now declared in the data, the template looks them up instead of falling through, and a gate refuses any verdict outside the list.

Trove's title endpoint returns its gazettes under the key newspaper. Asking it for a gazette key returns nothing, so the first run of the script that measures Australian holdings reported 0 gazettes — and that was one commit from being published as a fact about Trove. It is a fact about the parser. The endpoint's own total is now compared with what the parser found and a disagreement refuses the build: newspaper 2,014 = total 2,014, gazette 38 = total 38.