glinia.org · v0.15.3
An indexed corpus of archives, maps, sources and terms for the Polish–Lithuanian borderlands — with an evidence ledger that will not promote a claim without corroboration, and a watchdog for the ways surname research goes wrong.
glinia — clay; the wet, workable ground. The root the name descends from.
Looking for your own people is a good thing to do. Nothing here assumes otherwise. But this particular ground has been used, repeatedly and by several parties, to argue that some people belonged and others did not — and the tools of genealogy were part of how those arguments were made. The rules below are not accusations. They are the places where careful people have gone wrong before, marked so you don't have to rediscover them.
Two errors sit on the same axis and this tool guards against both. One is reading a name as a claim to rank or nation. The other is reading a nation's worst acts onto every family that shares its language. Neither individuals nor peoples are what their labels say they are.
The practical consequence: this tool reports what records say and how well supported each claim is. It will not output an ancestry, an ethnicity, or a status, and nothing rises above tier 2 without a document you have actually read.
Paste a conclusion you are tempted to draw — or anything you are about to write down — and it will be checked against the rules above. Nothing leaves your browser.
Every incorporated source is scored on five weighted metrics plus a separate slant reading. The scores are structured judgements with published weights, not measurements — but because every cell and every weight is visible, you can disagree with a specific number rather than with a vibe. Sort the table to see what the rubric actually claims.
It measures evidential use for this particular search — establishing facts about a named person in five districts of the Grand Duchy. Nothing else. With Fire and Sword scores near the floor and is a masterpiece; the rubric is saying it cannot tell you who your great-great-grandfather was, which is true and is not a criticism of the book.
Slant is deliberately excluded from the composite and shown separately. Folding it in would let a bland useless source outrank a biased essential one. A high slant score is an instruction about how to read something, not a reason to discard it.
| Source | Composite | Prox | Gran | Cov | Acc | Ind | Slant |
|---|
Click a row for the scoring note. Composite = Proximity 0.25 + Granularity 0.25 + Coverage 0.20 + Access 0.15 + Independence 0.15.
Full public-domain texts, not links to catalogues — the standing supply you draw from when a register entry needs a world around it.
Almost every one of them was written because the thing it describes was being destroyed. Mickiewicz wrote in Paris exile after the November Uprising collapsed. Gloger catalogued under a tsarist censor whose stamp is printed in every volume. Kolberg recorded villages as they dissolved. Sienkiewicz wrote to repair a nation that had just lost 1863. Only Wolne Lektury was made in a country that exists.
That shared climate has a consistent effect. Salvage texts over-preserve what felt endangered, under-record what felt ordinary, and idealise by design when the purpose was morale rather than description. Each card below states its climate and its purpose before its contents, because knowing what a book was for is the difference between using it and being used by it.
Searchable with type:text and filterable by licence:pd, format:epub or purpose:.
Licences differ and matter only if you republish. pd is public domain outright; pd-us is Gutenberg's jurisdictional statement, not a universal one; mixed means the text is free but the editorial apparatus is CC BY-SA.
The parallel edition, wired in. Pick a Book to see its three titles and the episode summaries of all three texts side by side — then ask about it. Answers are grounded in this data and in the corpus above; the companion is instructed never to invent a translation.
This reads Mickiewicz's 1834 Polish, Biggs's 1885 blank verse and Noyes's 1917 prose — all public domain. It holds each Book's titles, episode summaries and line counts, not the poem itself; for the text, open the parallel EPUB.
It is a companion to a poem, not evidence about your family. Nothing it says can raise a claim in the lineage ledger.
The edition itself is a separate EPUB: Pan Tadeusz — A Reading Edition, v05.05.00. Forty-one chapters, restructured so the story comes first. Noyes's 1917 prose runs straight through; the 1834 Polish and Biggs's 1885 blank verse follow as endnotes, each linked from the Book it belongs to.
Each Book opens with an outline — themes, where the action stands in time, what to watch for — carries one anchored note inside it, and closes with a short account of where the three versions part in that Book. Twelve of each.
The twelve notes are anchored to real occurrences — each attaches to the first appearance of a term verified to exist in that Book of Noyes's text, and lands after the paragraph containing it rather than inside one. Each says what a Polish institution was, how the two translators diverge over it, and where it applies, what it means for the search.
Built to survive readers that strip styling: every verse line is a real paragraph rather than a styled span, and no line break, table cell or hidden label is ever the only thing separating two words. Note markers use epub:type="noteref" and page boundaries epub:type="pagebreak" with aria-hidden — both more conformant than the sources it was built from. Filenames read as chapter titles, for readers that ignore the navigation document. It reads correctly with the stylesheet switched off entirely.
Every community below lived on the same ground your family did, in the same centuries, often in the same parishes. Each has its own archives and its own descendants searching them. This is here so the tool describes a shared place rather than an empty one waiting for your line — and so that if you find something belonging to someone else's search, you know where to send it.
Nothing in this section implicates your family in anything. It is the opposite: a reminder that the people who lived here were neighbours to each other, and that most of them left no more record than yours did.
Every claim carries a tier. Tiers are not opinions — they are computed from how many independent sources support the claim and whether a primary document has been cited. A claim cannot reach Verified without an archive reference you can point at.
Zero verified facts about your family exist in this ledger. Everything gathered so far is about the surname in general — its morphology, its distribution, its likely region of formation. None of it is about your ancestors, because no record bearing your ancestors' names has been read yet.
That is the honest starting position and the ledger will keep saying so until a document changes it. Claims marked about the name are general scholarship. Only claims marked about your line count as biography.
Add a claim with its sources. Tier is assigned automatically: one source is Single-source, two independent sources is Corroborated, and a cited primary document is Verified.
Tier will be computed on save.
Free text searches every field. Prefix type: tag: year: from: to: licence: score: slant: to filter, or "quote a phrase". Press / from anywhere. Every source — text, map or repository — carries its climate and its purpose, shown before its contents. Records carrying a caution show it in red.
Nine states of the same ground. Borders are schematic; the data layers are real. Each period can push its date range into the corpus search.
No slips yet.
Page images of about 4,000 Roman Catholic register books, with no surname index — you search by parish and read. Ejszyszki is online.
Genitive forms follow the documented pattern. Verify against the site's own spelling — a wrong inflection returns zero and looks exactly like absence.
Findings enter the corpus and are scanned against the integrity rules as you type. Search with type:finding. Stored on this device only.
A null in Podlaskie is evidence. A null in Litwa is a statement about transcription volunteers.
| Region | Code | Births | Deaths | Marriages | Read a null as |
|---|
GLINIA v0.15.3 · glinia.org · single file, no build step
Corpus indexed in-browser. Claims and findings stored locally, never transmitted. Borders schematic — verify figures before citing.