The archive holds scans; the records moved to the corpus

880,483 files leave this repository and none of them was a scan. 840,810 pages
of OCR, 38,725 bibliographic records, 1,743 files of plain text, the design
specs, the schemas and the `scans` tool itself — all of it describes the
facsimiles rather than being one, and all of it is now in the corpus that cites
this archive, at github.com/peasanttide/history: the records in data/, the text
in sources/, the tool in scans-tool/, the specs in docs/archive/.

Nothing was deleted that had not been verified present there first, file by
file: 38,725 of 38,725 records and 840,810 of 840,810 OCR pages. And nothing is
gone in any case — it is all in this repository's own history.

What remains is 39,324 PDFs and 94 JP2 map sheets, 137 GB, and a README that
describes them.

The split is worth stating rather than assuming. A facsimile is stable and
large; a record about it is small and gets corrected — the Hardy record was
corrected three times today. Keeping them apart means this can be mirrored,
pinned and left alone, while the description of it moves as often as it needs
to. It also means the addressing had to stop depending on where a file sits, and
it does: ids are flat, the directories here are cosmetic, and a citation
survives any reorganisation of them.

New in this commit: sources/marat, 931 PDFs and 38 GB — every surviving number
of Marat's four papers, 1789 to 1793, one PDF per issue, bound at native pixels
from the ARTFL Project's photographs of the New York Public Library set, with
Gallica supplying seven issues and one whole number that ARTFL's image server
does not have.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KNRQDxd9kCdeWWfgdLGXki

.claude/skills/transcribe-document/SKILL.md

@@ -1,302 +0,0 @@
----
-name: transcribe-document
-description: Use when transcribing a scanned document in this archive into markdown - a Journal de Paris issue or supplement, a bound insert, a map sheet. Covers rendering the pages from the PDF, transcribing each page from the image with headings and tables preserved, proofreading, marking what could not be read, and recording the result as [[page]] and [[page.text]] entries in the document's .toml.
----
-
-# Transcribing a document
-
-You are producing a **faithful, readable transcription of what is printed on the paper**, one
-file per page, cited from the document's record.
-
-Transcribe from the **page image**. Do not paste the PDF's text layer: it is a bad OCR pass of
-eighteenth-century type and it invents things. `NUMÉRO 5` is in there as `NUMERO 25`, and one
-page reads `NUMERO 807179`. It is useful for finding a page and worthless for quoting one.
-
-## What you produce
-
-For a document `<id>` with N pages, in the directory the record already lives in:
-
-```
-<id>.p1.md … <id>.pN.md      one transcription per page
-<id>.toml                    gains one [[page]] per page, each citing its .md
-```
-
-Nothing else changes. The markdown is never inlined into the TOML — the design spec settles
-that ("Text and OCR ... never inline in TOML"), and it is what keeps the record diffable.
-
-## The loop
-
-### 1. Find the document
-
-```bash
-./tools/target/release/scans.exe resolve <id>
-./tools/target/release/scans.exe resolve <id>.p1
-```
-
-The first prints the record's path and layer; the second adds the container and the page
-inside it, which tells you the addressing is sound before you spend anything on it. If the
-binary is missing, build it — `--release`, because a debug build is roughly thirty times
-slower and you are about to rasterise several pages:
-
-```bash
-cd tools && cargo build --release --features render
-```
-
-### 2. Render every page of the document, all three ways, before transcribing any of them
-
-```bash
-./tools/target/release/scans.exe render <id>
-./tools/target/release/scans.exe render <id> --grid 2x2
-./tools/target/release/scans.exe render <id> --grid 4x4
-```
-
-All of it lands in `.render/`, which is gitignored. Re-running is cheap: anything already
-rendered is left alone.
-
-| what you get | what it is for |
-|---|---|
-| `<id>.p<n>.png` — the whole page | the **layout**: how many columns, where the rules are, which lines are headings, what order to read in |
-| `<id>.p<n>.r<r>c<c>.png` at `2x2` — four tiles | **reading prose**. For a two-column page the vertical boundary falls near the gutter, so `r1c1`+`r2c1` is roughly the left column and `r1c2`+`r2c2` the right |
-| the same at `4x4` — sixteen tiles | **anything you must get exactly right**: every numeral, table cells, the printed page number, a word you are not sure of |
-
-Render all pages up front because **a word broken across a page is rejoined on the page where
-it starts**, and the rest of it is on the next page's image. When a page ends mid-word, open
-the next page's first tile, take the rest of the word, and finish the sentence on the page
-that began it. Nothing is left on the following page.
-
-You will be tempted to read numbers off the whole page. Do not. A whole page is about
-3500x4900 and arrives sampled down until the type is a few pixels tall — display type and
-table rules survive that, numerals do not. **Every digit you write down comes from a `4x4`
-tile.** That is what it is there for, and misreading `28. 4, 0` as `28. 4. 6` is the failure
-this transcription is most likely to contain.
-
-### 3. Transcribe, one page at a time
-
-Whole page for the shape of it, `2x2` tiles for the words, `4x4` tiles for anything exact.
-Write `<id>.p<n>.md` beside the record. No front matter, no title you invented, no page number
-heading — the file is the page, and the record says which page it is.
-
-### 4. Proofread it
-
-Read the tiles again against what you wrote. You are looking for the mistakes this task
-actually produces:
-
-- a word you guessed at from its shape rather than read
-- `f` where the paper has a long s, and the reverse
-- dropped accents, and accents added where the paper has none
-- numbers in tables transposed or shifted a column — check these at `4x4`
-- a line skipped where a tile boundary fell
-
-Fix what you can. Where you cannot, mark it — see *Marking what you could not read*.
-
-### 5. When every page is done, edit the record once
-
-One edit to `<id>.toml`. See *Recording it in the .toml*.
-
-### 6. Validate
-
-```bash
-./tools/target/release/scans.exe validate source/<path-to>/<id>.toml
-```
-
-Zero errors, zero warnings. `E706` means a `[[page.text]]` names a file that is not there —
-usually a typo in the filename, or a `.md` written in the wrong directory.
-
-## How to transcribe
-
-### Transcribe, do not translate and do not modernise
-
-The text stays in its own language and its own spelling. `connoître` stays `connoître`,
-`étoit` stays `étoit`, `Poëme` keeps its diaeresis, `enfans` keeps no `t`. `&` stays `&` and
-is not expanded to `et`. Æ and œ stay ligatured. Accents are as printed, including where
-eighteenth-century practice differs from modern French (`théatre`, `trés`).
-
-Three things are normalised, and only these three:
-
-- **The long s.** `ſ` is transcribed `s`. It is the same letter in a different sort, not a
-  different letter, and leaving it in makes the file unsearchable for no gain. `Ceſſer` is
-  written `Cesser`.
-- **Superscripts come down to the baseline**, keeping their letters: `Sr`, `Dlle`, `Dme`,
-  `Md`, `Cher`, `ve`, `Mgr`, and the ordinals `1er`, `2d`, `4e`, `83e`. `Nº` is written `No.`
-- **Spacing before punctuation is regularised** to one space before `;` `:` `?` `!`, and none
-  before `,` or `.` — which is French practice and what the paper is doing, but the paper's
-  spaces are thin and uneven and no two people would read them the same way twice.
-
-Editorial marks you add are in English and in square brackets, so they can never be mistaken
-for the document's own words.
-
-### Preserve the structure with markdown
-
-| on the page | in the file |
-|---|---|
-| the paper's masthead, first page of an issue only | `# JOURNAL DE PARIS.` |
-| the issue number above it, and the date line below | bold, each on its own line |
-| a department heading in large caps on a line of its own — `EXTRAITS.`, `ADMINISTRATION.`, `SPECTACLES.`, `VARIÉTÉS.` | `## EXTRAITS.` |
-| the title of a piece, on a line of its own, under one of those | `### BELLES-LETTRES.`, `### RÉSULTAT du Conseil d'État du Roi…` |
-| italic in the original — titles of works, cited matter, names in some settings | `*…*` |
-| a ruled table | a markdown table, its caption bold on the line above |
-| verse | a blockquote, one printed line per line |
-| a footnote keyed `(1)` at the foot of a column | at the foot of the page, after a `---` |
-
-Keep the printed capitalisation of headings.
-
-**A heading occupies a line of its own.** Caps at the head of a paragraph, with the text
-running on from them on the same line, are text — `THÉATRE FRANÇOIS.` run into its notice is
-part of the notice, while `PALAIS ROYAL.` centred on its own line is a heading. This is the
-distinction that decides it; nothing else does.
-
-**Heading levels record how the page sets the type, not a logical hierarchy.** `###` under a
-`##` does not claim the piece belongs to the department above it — a notice from the register
-of seals set as a titled piece is `###` even when the nearest `##` is `MUSIQUE.`, because the
-paper gives it the same rank as every other titled piece. Do not renumber levels to build a
-tree the page does not have.
-
-**Do not italicise inside a heading.** The heading level carries it. `### RÉSULTAT du Conseil
-d'État du Roi, tenu à Versailles le 27 Décembre 1788.`, not `### *RÉSULTAT…*`.
-
-**A rule gets a `---` only when it is doing work no heading is already doing.** The rule above
-a footnote does; the rule between two sections with nothing else to divide them does; the
-short ornamental rules between spectacle notices that each already carry a heading do not.
-
-Verse is a blockquote because it is the one construction where the **line breaks are the
-form**, and a blockquote keeps one printed line on one line of the file without depending on
-a renderer honouring trailing spaces:
-
-```markdown
-> O Monde ! aggrandis-toi : Copernic va paroître ;
-> Il paroît, il a dit, l'univers est changé.
-```
-
-A table cell the original merges down several rows — one `Vent.` reading standing against
-three `Époques.` — goes in the **first** of those rows, and the cells below it are left empty,
-however many columns are merged:
-
-```markdown
-| A 7 h. ½ m. | — 17, 4 | 28. 2, 9 | E. S. E. | Ciel assez beau toute la matinée. |
-| A 2 s. | — 9, 8 | 28. 1, 8 | | |
-| A 9 s. | — 9, 0 | 27. 11, 2 | | |
-```
-
-Markdown cannot merge cells, and repeating the value would state three observations where the
-paper made one. The caption goes bold on the line above **wherever the paper puts it** — the
-meteorological table's caption is set rotated in the left margin inside a brace, and it still
-goes above.
-
-**Prose line breaks are not preserved.** A paragraph is one line in the file. A word broken
-across a line by a hyphen is rejoined — `impor-` + `tante.` is `importante.`
-
-**Column breaks are not marked.** The text simply continues.
-
-### What is not part of the document
-
-Leave these out entirely:
-
-- the digitiser's watermark (`Digitized by Google`)
-- library stamps, shelfmarks, and accession numbers added by a holding institution
-- signature marks and catchwords at the foot of a leaf — binder's apparatus, not text
-- bleed-through from the other side of the leaf
-
-The **printed page number** is also not transcribed into the markdown. It goes in the record,
-as `label`. See below.
-
-### Marking what you could not read
-
-In the markdown, in place of the text:
-
-| mark | for |
-|---|---|
-| `[illegible]` | a word or more you cannot read at all |
-| `[illegible: 3 words]` | when you can count what is missing |
-| `[unclear: Sorin]` | you have a reading but would not stake a quote on it |
-| `[torn]`, `[stained]`, `[cropped]` | the page itself is damaged or the scan cuts it off |
-| `[sic]` | the paper's own error, transcribed as printed |
-
-An editorial mark **replaces the text and its formatting both**. An italic word you cannot be
-sure of is `[unclear: Abraham]`, not `*[unclear: Abraham]*` — the mark is your voice, and your
-voice is not italic.
-
-`[sic]` is the important one: **a misprint in the original is preserved, not corrected**. If
-the paper prints `fauhourg` for `faubourg`, or a date that cannot be right, transcribe it as
-it stands, mark it `[sic]`, and say so in the note. Silently fixing the source is the one
-thing a transcription must never do — a proofread fixes *your* reading of the page, never the
-page.
-
-Then summarise the page's unresolved problems in **one line** on that page's `[[page.text]]`
-`note`. Omit `note` entirely when there is nothing unresolved — an empty note is noise across
-eight hundred pages. Good notes:
-
-```toml
-note = "Foot of the second column is cropped by the scan; four lines are lost."
-note = "Prints \"fauhourg\" for \"faubourg\", transcribed as printed."
-note = "Thermometer readings in the table are faint; the second row is a best reading."
-```
-
-## Recording it in the .toml
-
-One `[[page]]` per page, in order, each with one `[[page.text]]`. Keep everything already in
-the file, `#:schema` first, and keep `pages = { from, to }` — the loader checks that the
-number of `[[page]]` entries equals the length of that range and errors if they disagree.
-
-```toml
-#:schema ../../../../../schemas/source.json
-id    = "journal-de-paris-1789-01-02"
-layer = "document"
-of    = "journal-de-paris-1789-vol1"
-type  = "issue"
-no    = 2
-date  = "1789-01-02"
-pages = { from = 9, to = 12 }
-
-[[page]]
-n     = 1
-label = "5"
-[[page.text]]
-file = "journal-de-paris-1789-01-02.p1.md"
-kind = "transcription"
-by   = "claude-opus-5"
-lang = "fr"
-
-[[page]]
-n     = 2
-label = "6"
-[[page.text]]
-file = "journal-de-paris-1789-01-02.p2.md"
-kind = "transcription"
-by   = "claude-opus-5"
-lang = "fr"
-note = "Second column is thickly inked; the last word of the page is a best reading."
-```
-
-- `n` runs `1..N`. It is the page's number **within this document**, which is what `.pN`
-  addresses.
-- **`label` is read off the page and is never calculated.** It is the folio the printer set,
-  as a string. It is *not* `pages.from`, and it is *not* `n`: `pages` counts pages of the
-  scanned PDF, which includes leaves the printer never numbered. The example above is real —
-  the 2 January issue occupies PDF pages 9 to 12 and is printed 5 to 8, because four
-  unnumbered leaves open the volume. Read the corner of each page at `4x4`. If you cannot read
-  it, omit `label` and say so in the note; do not guess it from the page before.
-- `kind = "transcription"`, not `ocr`. `ocr` means a machine pass nobody read.
-- `by` names who produced it, because the archive expects to be re-transcribed one day and
-  needs to know which pass a page is still on.
-- Follow the file's existing formatting: `=` aligned within each block, `[[page]]` and
-  `[[page.text]]` flush left, keys in the order shown.
-
-## A worked page
-
-`example-page.md`, beside this file, is the first page of the first issue of 1789
-(`journal-de-paris-1789-01-01.p1`) transcribed to these rules. Read it before you start. It
-shows, on one page: the masthead, a caption plus a ruled table with merged cells, two heading
-levels, a long paragraph reassembled across a column break with its hyphenation rejoined,
-italics on work titles, verse as a blockquote, and a keyed footnote below a rule.
-
-## Do not
-
-- Do not touch `id`, `of`, `layer`, `type`, `no`, `date`, `pages`, or `supplement_to`. If one
-  of them looks wrong, say so in your report and leave it alone — those come from the ingest,
-  and correcting one by hand hides a bug in it.
-- Do not transcribe a page you did not look at.
-- Do not summarise, abridge, or "clean up" the text. Every word on the page is in the file.
-- Do not skip a blank page. It still gets a `[[page]]`, or the count stops matching the range.
-  Its file holds the single line `[blank page]`, and its text entry carries
-  `note = "Blank page."` — an absent file is a validation error, and an empty one is
-  indistinguishable from a transcription that was never written.

.claude/skills/transcribe-document/example-page.md

@@ -1,41 +0,0 @@
-**NUMÉRO 1er**
-
-# JOURNAL DE PARIS.
-
-**Jeudi 1er JANVIER 1789, de la Lune le 6.**
-
-LE SOLEIL se lève à 7 heures 52 minut., & se couche à 4 heures 8 minut.
-
-LA LUNE se lève à 9 h. 55 min. du matin, & se couche à 9 h. 36 min. du soir.
-
-*Rapport du Tems vrai au Tems moyen.* Au midi du Soleil, la pendule doit marq. 0 h. 4 m. 22 s.
-
-*Hauteur de la Rivière.* Le 30 à 0 p. 9 p., & le 31 à 1 p. 0 p. (haut. moyenne 5 pieds.)
-
-*Réverbères.* Allumés à 5 heur. 0 min., éteints à 3 h. 0 min.
-
-**Observations Météorologiques du 30 Déc.**
-
-| Époques. | Thermom. | Baromèt. | Vent. | État du Ciel & Remarques. |
-| --- | --- | --- | --- | --- |
-| A 7 h. ½ m. | — 14, 6 | 28. 4, 0 | Calme. | Ciel brumeux dans la matinée, pur le reste de la journée. |
-| A 2 s. | — 10, 4 | 28. 4, 0 | | |
-| A 8 ¾ s. | — 12, 9 | 28. 4, 0 | | |
-
-Dans la nuit du 30 au 31, à 5 h. ¼ du mat., le therm. à 17 dég. 3 dixes au dessous de la glace.
-
-## EXTRAITS.
-
-### BELLES-LETTRES.
-
-*ALMANACH DES MUSES*, 1789. A Paris, chez *Delalain* l'aîné & fils, Libraire, rue St-Jacques, No. 240. In-12 d'environ 300 pag. ; prix 1 liv. 16 s.
-
-La plupart des esprits, pendant l'année 1788, se sont tournés vers des objets fort étrangers à la Poésie ; aussi la Notice des Ouvrages de Poésie qui termine *l'Almanach des Muses* est-elle beaucoup moins considérable qu'à l'ordinaire. Cependant le Recueil lui-même se ressent moins de cette stérilité. Il contient un grand nombre de Pièces importantes par leur étendue & par le talent qui les a produites, telles que *les États-Généraux de Cythère*, Imitation très élégante du *Congrès de Cythère* du Comte *Algarotti*, par M. le Cher *de Cubières* ; le premier Chant d'un Poëme de *Vulcain*, sur lequel M. *Pons de Verdun* a répandu beaucoup d'enjouement & d'esprit, & une grande fécondité d'imagination ; *mon Retour à Paris*, par M. le Cher *Dupuy des Islets*, qui y a rassemblé une suite de tableaux plus piquans les uns que les autres, & y a développé un talent très distingué dans l'art des vers ; *les Adieux d'un Gouverneur à son fils*, par M. *Marichal*, morceau rempli d'excellens préceptes ; une Ode pleine de noblesse & d'harmonie adressée à Mme *St-Hubert* par M. *Sorin*, qui l'a louée d'une manière digne de son talent, en évitant, avec grand soin, les lieux communs & les exclamations prétendues lyriques ; *La Construction des Hôpitaux*, Poëme de M. *Theveneau*, dont la manière paroît avoir gagné sensiblement ; *Alcippe & Lindor*, par M. *Alix*, Dialogue dans lequel un Misantrope répand les traits d'une causticité piquante sur les mœurs & les travers du siècle, &c. Mais le morceau qui, selon nous, sera le plus remarqué, tant par la grandeur du sujet que par l'élévation du ton que le Poëte y a pris, est un *Essai sur l'Astronomie*, par M. *de Fontanes* (1). Ce titre d'*Essai* est modeste : car cet Ouvrage est un véritable Poëme. L'Auteur y expose les divers systêmes astronomiques anciens & modernes, jusqu'à celui qui est universellement reconnu par les Savans pour le seul véritable. Ensuite son imagination prend ainsi l'essor à l'aspect du spectacle immense qui la frappe :
-
-> O Monde ! aggrandis-toi : Copernic va paroître ;
-> Il paroît, il a dit, l'univers est changé.
-> Seul, au centre du monde, à son poste rangé,
-
----
-
-(1) Cet Essai est tiré du Recueil en deux volumes, que M. *de Fontanes* doit publier incessamment.

.gitignore

@@ -12,3 +12,7 @@
 .push-hf.sh
 .hardy.log
 .hf-upload.log
+.staging/
+
+# Loose page images staged while a PDF is being bound.
+.staging/

.taplo.toml

@@ -1,53 +0,0 @@
-# Schema enforcement for the taplo CLI and CI.
-#
-# schemas/source.json is GENERATED from the Rust types in tools/src/model.rs by
-# `scans schema`. Never edit the schema by hand: `scans schema --check` fails when the
-# checked-in file has drifted away from the types the validator itself uses.
-#
-# SCOPE OF THIS FILE
-#
-# This config drives the taplo CLI (`taplo check`, `taplo fmt`). It does NOT drive the
-# VS Code extension: the bundled language server was observed loading this file and then
-# associating no schema from it in any form tested - relative or absolute include, root
-# level `[schema]` or `[[rule]]`. Editor validation comes from the `#:schema` directive
-# on line 1 of each record instead. See schemas/README.md.
-#
-# Every archive record therefore still carries its own directive, e.g.
-#
-#     #:schema ../../schemas/source.json
-#
-# and the rule below is the belt to that file's braces: it applies the same schema to
-# anything under source/ that is missing its directive, so a record cannot escape
-# validation in CI merely by forgetting the line.
-
-# Do not walk build output or the Rust crate; the schema describes archive records only,
-# and Cargo.toml is not one.
-exclude = ["target/**", "tools/target/**", "tools/Cargo.toml", "node_modules/**"]
-
-# Archive records: everything under source/ is a source, copy or document record. The
-# `layer` key picks which of the three, and the schema's oneOf narrows completions and
-# the required-field set accordingly.
-#
-# NOTE: include globs are resolved relative to this file's directory. On Windows this
-# resolution has been seen to produce mixed path separators that stop the glob matching;
-# if `taplo check` reports no files on Windows, run it from the repo root or pass the
-# paths explicitly:  taplo check "source/**/*.toml"
-[[rule]]
-name = "archive-record"
-include = ["source/**/*.toml"]
-
-[rule.schema]
-path = "schemas/source.json"
-
-# Formatting, kept identical to the evenBetterToml.formatter.* settings in
-# .vscode/settings.json so the editor and the CLI cannot disagree.
-#
-# reorder_keys is off on purpose: records are written in a deliberate order - identity
-# (`layer`, `id`, `of`) first, then description, then the [[page]] array last - and
-# alphabetising them would destroy that reading order.
-[formatting]
-align_entries = true
-array_trailing_comma = true
-compact_arrays = false
-indent_string = "  "
-reorder_keys = false

.vscode/extensions.json

@@ -1,9 +0,0 @@
-{
-  // Even Better TOML provides the TOML language server (taplo) that reads each
-  // record's `#:schema` directive and validates it against schemas/source.json.
-  // Without it the archive's TOML files get no schema validation in the editor.
-  //
-  // rust-analyzer is for the `scans` tool in tools/, which generates that schema.
-  "recommendations": ["tamasfe.even-better-toml", "rust-lang.rust-analyzer"],
-  "unwantedRecommendations": []
-}

.vscode/settings.json

@@ -1,60 +0,0 @@
-{
-  // ---------------------------------------------------------------------------
-  // Schema enforcement for the editor.
-  //
-  // Requires the "Even Better TOML" extension (tamasfe.even-better-toml); see
-  // .vscode/extensions.json. schemas/source.json is GENERATED from the Rust types
-  // in tools/src/model.rs by `scans schema`. Never edit it by hand -
-  // `scans schema --check` fails when it has drifted.
-  //
-  // Because the schema is a oneOf discriminated on `layer`, typing
-  // `layer = "copy"` narrows both autocomplete and the required-field set to the
-  // copy layer alone. Verified: a source without `title` is flagged, a document
-  // without `title` is not, and `layer = "page"` is rejected outright.
-  //
-  // HOW RECORDS GET THE SCHEMA
-  //
-  // Each archive record carries its own `#:schema` directive on line 1, pointing
-  // at schemas/source.json relative to that file, e.g. from source/turgot/:
-  //
-  //     #:schema ../../schemas/source.json
-  //
-  // The directive is the ONLY portable mechanism. It is committed with the data,
-  // needs no per-machine setup, works for every contributor, and is honoured by
-  // both this extension and the taplo CLI.
-  //
-  // `evenBetterToml.schema.associations` is deliberately NOT used here, and the
-  // reason is not style. Its values must be absolute URIs with an http/https/file
-  // scheme. A repo-relative value such as "./schemas/source.json" is resolved
-  // against an internal `root:///` base, not the workspace, so the schema silently
-  // fails to load: the association matches, no diagnostics are ever produced, and
-  // the editor reports nothing at all. `${workspaceFolder}` is NOT expanded, and a
-  // bare Windows path is parsed as a URL with the scheme `c`. All three were tested
-  // against the bundled language server; every one of them yields zero validation
-  // while looking correct. Do not add one back.
-  //
-  // If you want a machine-local association anyway, it must be an absolute file
-  // URI, and it belongs in your USER settings, never in this committed file:
-  //
-  //   "evenBetterToml.schema.associations": {
-  //     ".*/source/.*\\.toml$": "file:///d:/source/scans/schemas/source.json"
-  //   }
-  // ---------------------------------------------------------------------------
-
-  "evenBetterToml.schema.enabled": true,
-  "evenBetterToml.schema.links": true,
-
-  // Keep the formatter in step with [formatting] in .taplo.toml, so the editor and
-  // the CLI cannot disagree about how a record is laid out.
-  "evenBetterToml.formatter.alignEntries": true,
-  "evenBetterToml.formatter.reorderKeys": false,
-  "evenBetterToml.formatter.indentString": "  ",
-  "evenBetterToml.formatter.arrayTrailingComma": true,
-  "evenBetterToml.formatter.compactArrays": false,
-
-  "[toml]": {
-    "editor.defaultFormatter": "tamasfe.even-better-toml"
-  },
-
-  "rust-analyzer.linkedProjects": ["tools/Cargo.toml"]
-}

README.md

@@ -6,9 +6,9 @@ language:
   - en
 tags:
   - french-revolution
-  - ocr
   - primary-sources
   - digital-humanities
+  - facsimile
 pretty_name: Scans — primary sources of the French Revolution
 size_categories:
   - 10K<n<100K
@@ -16,307 +16,76 @@ size_categories:
 
 # scans
 
-Scans of primary sources from the French Revolution.
+Facsimiles of primary sources of the French Revolution. **Scans, and nothing
+else** — 39,324 PDFs and 94 JP2 map sheets, 137 GB.
 
-The product of this repository is **stable, citable addresses**. Structured facts and
-translations live elsewhere and cite into this one. Everything here exists to make those
-addresses resolvable, unambiguous, and correct.
+The records that describe these files, the transcriptions and OCR of what is
+printed on them, and the tool that fetched and bound them all live in the corpus
+that cites this archive: <https://github.com/peasanttide/history>. This
+repository is the bytes.
 
-## Addresses
+That split is the point. A facsimile is stable and large; a record about it is
+small and gets corrected. Keeping them apart means the archive can be mirrored,
+pinned and left alone, while the description of it moves as often as it needs to.
 
-```
-turgot-1739                     a record
-turgot-1739.p0                  its first page — the atlas key sheet
-journal-de-paris-1789-01-03.p1  page 1 of that day's issue
-```
-
-Ids are flat and never contain `/`, so **directories are cosmetic**: folders can be
-reorganised without breaking a single citation. Ask the tool what an address points at:
-
-```sh
-cd tools && cargo run -- resolve turgot-1739.p0
-```
-
-## The model
-
-Four layers, named after their TEI equivalents. `layer` says which; `type` says the genre
-within it (`newspaper`, `map`, `issue`, `supplement`, `engraving`, …).
-
-| `layer`    | what it is                                        | example                      |
-| ---------- | ------------------------------------------------- | ---------------------------- |
-| `source`   | the publication or work, independent of any copy  | `journal-de-paris`           |
-| `copy`     | one physical object that was digitised            | `journal-de-paris-1789-vol1` |
-| `document` | the citable intellectual unit                     | `journal-de-paris-1789-01-03`|
-| `page`     | one scanned side, carrying graphics               | `…-01-03.p1`                 |
-
-Layers may be skipped: `of` points at any ancestor, so a one-off engraving is a single
-`document` file and does not pay four files of ceremony. A page is never a file of its own —
-it is declared inline as `[[page]]` in the record that owns it.
-
-A file states only what is new. `[rights]`, `[[resp]]`, `language`, `scan.file` and the rest
-are inherited down the `of` chain; `date`, `title`, `note`, `url` and `[[page]]` are not.
-
-## Layout
-
-```
-sources/
-  frc/                                      the Newberry French Revolution Collection
-    ab/abondance00unse/
-      abondance00unse.toml                  source, one per Internet Archive item
-      abondance00unse.p1.ocr.md             OCR, one file per page
-      abondance00unse.pdf                   the scan
-    …38,377 items across ~700 two-character shards
-  journal-de-paris/
-    journal-de-paris.toml                   source
-    1789/
-      journal-de-paris-1789-vol1.pdf
-      journal-de-paris-1789-vol1.toml       copy
-      journal-de-paris-1789-vol2.pdf
-      journal-de-paris-1789-vol2.toml       copy
-  hardy/
-    hardy.toml                              source, the eight-volume manuscript
-    hardy-vol1.toml                         copy, 658 pages inline as [[page]]
-    hardy-vol1.pdf                          the bound volume; pages address into it
-    …hardy-vol2 … hardy-vol8
-    wikimedia/                              GENERATED by `scans wikimedia hardy`
-      hardy-1-work.qs … hardy-3-links.qs    QuickStatements, staged
-      hardy-vol1.commons.wiki               one Commons file page per volume
-  turgot/
-    turgot-1739.toml                        source, 21 sheets inline as [[page]]
-    turgot-sheet00.jp2 … turgot-sheet20.jp2
-  verniquet/
-    verniquet-1795.toml                     source, 73 sheets inline
-    verniquet-sheet00.jp2 … verniquet-sheet72.jp2
-  jefferson-to-jay-1789-06-17/
-    jefferson-to-jay-1789-06-17.toml        source, one letter, 7 leaves inline
-    jefferson-to-jay-1789-06-17.pdf         the bound leaves
-schemas/source.json                         GENERATED from tools/src/model.rs
-schemas/ocr.json                            GENERATED from tools/src/ocr.rs
-tools/                                      the `scans` CLI
-docs/                                       the design spec
-```
-
-Binary assets are in Git LFS.
+## What is here
 
-## The tool
+| collection | files | size | what it is |
+|---|---:|---:|---|
+| `frc` | 38,377 pdf | 78 GB | the Newberry Library's French Revolution Collection, digitised by the Internet Archive. One PDF per item, sharded two deep by the item's Internet Archive identifier. |
+| `marat` | 931 pdf | 38 GB | every surviving number of Jean-Paul Marat's four papers, 1789–1793, one PDF per issue. Page images from the ARTFL Project at the University of Chicago, photographed from the New York Public Library's seven-volume set; seven issues and one whole number come from Gallica, which has leaves ARTFL's server does not. |
+| `hardy` | 8 pdf | 18 GB | Siméon-Prosper Hardy's autograph journal *Mes loisirs*, 1764–1789. Eight manuscript volumes, 4,227 leaves, from Gallica (BnF, Français 6680–6687). |
+| `verniquet` | 73 jp2 | 2.3 GB | the *Atlas du plan général de la ville de Paris*, 1795. |
+| `turgot` | 21 jp2 | 1.2 GB | the Plan de Turgot, 1739, from the David Rumsey Map Collection. |
+| `journal-de-paris` | 2 pdf | 205 MB | the 1789 run of the first daily newspaper in France, two bound volumes. |
+| `actes-des-apotres` | 4 pdf | 116 MB | *Les Actes des Apôtres*, nos. 1–119, four volumes. |
+| `pellet-1873` | 1 pdf | 11 MB | Marcellin Pellet's 1873 study of it. |
+| `jefferson-to-jay-1789-06-17` | 1 pdf | 2.1 MB | Jefferson's despatch of 17 June 1789, seven leaves. |
 
-The `scans` CLI is **not in this dataset** — it lives in the source repository at
-<https://github.com/peasanttide/scans>. What is here is the archive it produces: the records,
-the OCR, and the page images. Everything below describes what that tool does, for readers who
-want to regenerate or extend the archive.
+## Addressing a page
 
-`tools/` in the source repository is a Rust crate producing the `scans` binary. One definition in
-`tools/src/model.rs` drives three things at once: Rust type checking, TOML deserialisation,
-and — via `schemars` — `schemas/source.json`. The schema is never hand-written, or it would
-drift from the validator and start lying.
+A file is addressed by the record that names it, over in the corpus. Ids there
+are flat and never contain `/`, so **the directories here are cosmetic** and can
+be reorganised without breaking a single citation:
 
-```sh
-cd tools
-cargo run -- validate            # the ten semantic checks
-cargo run -- validate --strict   # warnings fail the build too
-cargo run -- resolve <address>   # what does this citation point at?
-cargo run -- schema --check      # has the schema drifted from the types?
-cargo test
 ```
-
-`validate` never opens an image by default — the archive holds several gigabytes of JPEG
-2000. The checks that read bytes (real image dimensions, real PDF page counts) are behind an
-optional feature:
-
-```sh
-cargo run --features probe -- validate --probe
+lami-du-peuple-1790-07-01.p3     page 3 of that issue
+turgot-1739.p0                   the atlas key sheet
 ```
 
-## Pixels
-
-Two commands turn a citation into an image. Both take an address, so nothing ever names a
-page number inside a PDF by hand.
+Ask the corpus what an address points at, and what is printed on it:
 
 ```sh
-cargo run --release --features render -- render journal-de-paris-1789-01-03
-cargo run --release --features render -- render journal-de-paris-1789-01-03.p1 --grid 2x2
-cargo run --release --features render -- extract journal-de-paris-1789-01-03.p1
+history show lami-du-peuple-1790-07-01 --page 3
 ```
 
-`render` rasterises the page. `extract` writes out the bitmaps stored on it, untouched.
-Output goes to `.render/` at the repository root, which is gitignored: these are derived
-files, reproducible at any time, and at native resolution they outweigh the PDFs.
+## How the PDFs are made
 
-* **Resolution is native by default** — the scan's own, taken from its largest embedded
-  image. Journal de Paris pages come out around 3500x4900. `--dpi` overrides it.
-* **`--grid RxC` cuts the page into tiles** with `--overlap` percent of slack on the interior
-  edges. A whole page shown to a reader that samples it down to fit is illegible type; for
-  this paper's two-column setting, `--grid 2x2` puts roughly one half-column in each tile.
-* Both are behind the `render` feature, which is what pulls in the rasteriser. Build
-  `--release`: a debug build is about thirty times slower, and it shows over 888 pages.
+Every page is one `/DCTDecode` XObject holding the JPEG exactly as the library
+served it — nothing resampled, requantised or colour-converted, because the point
+of an archival scan is that it is the bytes the holding institution published.
+The page box comes from the image's own JFIF density, so a leaf shot at a
+different size stays a different size rather than being normalised into a
+uniformity the object does not have. Page order comes from the record, never from
+sorting a directory: the record is what gets corrected when a leaf turns out to
+have been shot twice or out of sequence.
 
-`hayro` does the decoding rather than `lopdf`, because every page image in volume 1 is JBIG2
-and volume 2 adds JPEG 2000 plates. It is pure Rust, so there is no Ghostscript, pdfium or C
-toolchain to install.
+Where a source serves progressive JPEG — Gallica does — each stream is re-emitted
+with `jpegtran -copy all`. `-copy none` would drop the APP0 density and every page
+box would become a guess.
 
-Each record carries a `#:schema` directive on its first line, which is what gets it validated
-in an editor. See `schemas/README.md` — the alternative, an editor schema association, was
-tested and silently does nothing.
+Linearisation is skipped above 2 GB. A linearisation table states byte offsets,
+and past `i32::MAX` the decoders disagree about what they are looking at: one
+reports a 2.5 GB volume as a single page, another emits twenty-one warnings per
+open. Two of the Hardy volumes are affected.
 
-## Binding
-
-`render` and `extract` take pages out of a container. `pdf` goes the other way: it binds a
-record's loose page scans into one, embedding the images without touching a pixel.
-
-```sh
-cargo run --release --features pdf -- pdf hardy-vol1
-cargo run --release --features pdf -- pdf hardy          # every copy under the source
-```
+## Rights
 
-Page order comes from the record's `[[page]]` array, never from sorting the directory — a
-record is the thing that gets corrected when a leaf turns out to have been shot out of
-sequence, and a directory listing would ignore the correction. Page size comes from the
-image's own JFIF density, so the object states the physical size it really has.
+The works are out of copyright. The digitisations are not all offered on the same
+terms, and the corpus records what each holding institution claims, unchanged,
+rather than rewriting it to match what Wikimedia Commons would accept. Gallica
+asserts a non-commercial restriction over its images; the Internet Archive's
+Newberry scans carry no such claim. The `[[copyright]]` block on a record says
+which applies to it.
 
-Gallica serves **progressive** JPEG and PDF's `DCTDecode` is specified against baseline, so
-each stream is transcoded through `jpegtran`, which re-emits the existing DCT coefficients
-sequentially. That is lossless, not a re-compression: no requantisation, no generation loss,
-identical decoded pixels. The claim is checked rather than trusted — every page is decoded on
-both sides with `zune-jpeg` and compared, and a single differing byte fails the build.
-
-The finished file is then **linearised** — first page's objects, a partial cross-reference
-table and a hint stream moved to the front, so a reader speaking HTTP byte ranges shows page
-one after a few megabytes instead of fetching the whole file. `qpdf` does that, not
-Ghostscript's `-dFastWebView`: Ghostscript reaches it through `pdfwrite`, which re-encodes
-images on the way past and would undo the whole point. Every `DCTDecode` stream is read out
-of both files afterwards and compared byte for byte. `--no-linearize` skips it.
-
-**Not above 2 GB, though**, and Hardy's volumes are all past it. A linearization table states
-its offsets as byte positions, and beyond `i32::MAX` those stop fitting in the 32-bit integer
-some readers hold them in. Measured on volume 1 at 2.55 GB: linearised, `hayro` — the decoder
-this crate's own `render` and `extract` use — reports the file as **one page**, and poppler
-emits 21 warnings per open before recovering through the main cross-reference table.
-Ghostscript is unaffected and `qpdf --check` calls the file correct, so the table is well
-formed; it is simply too big for those readers. The same qpdf invocation on a 12.8 MB slice
-of the same volume upsets nobody, which is what pins the cause to size.
-
-So above 2 GB the build says why and leaves the file alone, rather than shipping a volume
-that loads its first page quickly and cannot be opened by the tool that reads the archive.
-
-`jpegtran` and `qpdf` are operator tools, like `git-lfs`: install `libjpeg-turbo-progs`
-(Debian), `libjpeg-turbo-utils` (Fedora) or `jpeg-turbo` (Homebrew), and `qpdf`, which
-everyone calls `qpdf`. Neither is linked against, so the crate stays pure Rust.
-
-## Wikimedia
-
-`wikimedia` turns a source and the copies beneath it into the edits somebody would otherwise
-type by hand. It opens no socket and uploads nothing: the output is files, to be read and
-then pasted.
-
-```sh
-cargo run -- wikimedia hardy
-cargo run -- wikimedia hardy --work Q12345678    # once stage 1 has been run
-```
-
-Three QuickStatements batches and one Commons file-description page per copy, written to
-`wikimedia/` beside the source record. The batches are staged because QuickStatements v1 can
-address a new item only as `LAST`, so one batch cannot both create a work and create the
-volumes that point at it; `$WORK` and `$VOLn` mark where a Q-id belongs. `--work` also
-rewrites stage 1 to address that item rather than create one, so it can be run again to
-finish an attempt that failed part way through: QuickStatements answers `Statement already
-exists` for what is there and adds the rest.
-
-Every property and item id lives in `wikimedia::vocab` with the label it carried when it was
-verified, and every generated statement prints that label in a trailing `/* … */` comment —
-QuickStatements' own comment syntax, which it strips and reuses as the edit summary. Property
-numbers are not memorable and are easy to get subtly wrong �� `P7101` is *privacy policy URL*,
-not *Gallica ID* — so a mistake shows up in review as a wrong word rather than a wrong number.
-
-## OCR
-
-Word-level OCR lives in a sidecar beside the record it belongs to, one markdown file per
-scanned page, reached through the `[[text]]` that points at it.
-
-```markdown
----
-of: 1789iemilseptcen00unse
-page: 6
-engine: ABBYY FineReader 11.0
-lang: fr
-w: 2597
-h: 4418
-dpi: 500
----
-
-<the page's text>
-```
-
-Frontmatter says which page of what, and how big the image is. The body is the page. That is
-the whole format.
-
-**Scalars are bare unless they would lie.** `lang: fr`, not `lang: "fr"` — the frontmatter is
-meant to read as YAML. But `no` is the ISO 639-1 code for Norwegian and, to every YAML 1.1
-parser, the boolean false; a Python user loading that with PyYAML gets `False`. So a value is
-written bare when it round-trips as itself and quoted when it would not, which keeps the
-common case clean and the pathological case correct.
-
-**What is deliberately not here:** word boxes, per-word confidence, baselines, and DjVu's
-nesting. An earlier version carried all of it and could write the source XML back out
-unchanged. The boxes were seven eighths of the bytes — 8.3 GB against about 1.2 GB for the
-text — and were not being used. Nothing is destroyed by that: the coordinates are still in
-`XML_for_OCR/` in [frc-data](https://github.com/NewberryDIS/frc-data), one clone away, so a
-later version that wants them can have them without asking archive.org for anything.
-
-```sh
-cargo run -- ocr import x_djvu.xml --out ./pages
-cargo run -- ocr check sources/frc/ab/abondance00unse
-```
-
-## The French Revolution Collection
-
-38,377 Internet Archive items from the Newberry Library, listed by
-[NewberryDIS/frc-data](https://github.com/NewberryDIS/frc-data). Records and OCR come from a
-local clone of that repository; only the PDFs are fetched.
-
-```sh
-git clone --filter=blob:none --no-checkout https://github.com/NewberryDIS/frc-data.git
-cd frc-data && git sparse-checkout set --cone Metadata XML_for_OCR && git checkout master
-
-cargo run --release -- frc ingest --frc-data ../frc-data      # records + OCR, ~3 minutes
-cargo run --release --features fetch -- frc fetch             # the PDFs, ~21 hours
-```
-
-`frc fetch` makes one request at a time, a second apart, with a `User-Agent` naming this
-repository, and verifies every download against the size and MD5 archive.org publishes for
-it. It resumes from what is already on disk, so it can be killed at any moment.
-
-## Licence
-
-The **database** — the records, the OCR sidecars, the structure and the addresses — is under
-the [Open Database License](https://opendatacommons.org/licenses/odbl/) (ODbL 1.0). Use it,
-adapt it, build on it; if you publish a derived database, share it alike and keep the
-attribution.
-
-The **contents are a separate question, and mostly not ours to license.** Every work here is
-an 18th-century imprint or a 19th-century study of one, long out of copyright everywhere, and
-the digitisations are published as public domain by the institutions that made them. Each
-record says so for itself:
-
-```toml
-[rights]
-work        = "PD-old-100-expired"
-scan        = "PD"
-attribution = "Digitised by the Internet Archive, sponsored by The Newberry Library."
-```
-
-So: quoting a pamphlet, reprinting a page image, or lifting one item's OCR carries no
-obligation from us — that material is public domain and stays public domain. ODbL bites on
-the *collection*: the selection, the identifiers, the metadata mapping and the structured OCR
-taken together. That distinction is the point of using a database licence rather than a
-content one, and it is why the per-record `[rights]` fields are not redundant with this
-section.
-
-Scans and metadata originate with the [Newberry Library](https://www.newberry.org/) and the
-[Internet Archive](https://archive.org/), and the item list comes from
-[NewberryDIS/frc-data](https://github.com/NewberryDIS/frc-data).
-
-## Design
-
-`docs/superpowers/specs/2026-08-05-source-archive-schema-design.md` is the authority: the
-four layers, the inheritance allowlist, EDTF dates, and the ten validation checks, with the
-reasoning for each.
+Binary assets are in Git LFS.

docs/superpowers/specs/2026-08-05-source-archive-schema-design.md

@@ -1,476 +0,0 @@
-# Primary source & scan archive — schema design
-
-Status: approved design, not yet implemented
-Date: 2026-08-05
-
-## Purpose
-
-This repository is an archive of primary sources from the French Revolution and the
-scans of them. Its product is **stable, citable addresses**. Structured facts and
-translations live in a separate repository, which cites into this one. Everything in
-this schema exists to make those addresses resolvable, unambiguous, and correct.
-
-The game consumes three things through those addresses: pixels (crops, map tiles),
-OCR'd text, and quotes with citations.
-
-## Diagnosis of the current layout
-
-The existing tree works but does not scale, for three specific reasons.
-
-**`kind` mixes three dimensions into one field.**
-
-| current `kind` value | what it actually says |
-|---|---|
-| `newspaper`, `map` | genre of the publication |
-| `volume` | this is a physical object |
-| `issue`, `supplement` | type of the intellectual unit |
-
-The chain of `of = ...` therefore looks uniform while meaning something different at
-every level.
-
-**A layer is missing.** The Lausanne bound volume — shelfmark 1094184846, scanned by
-Google, front matter trimmed — is not the same thing as "Journal de Paris, 1789,
-volume 1" the publication. One is paper in Switzerland; the other is an editorial
-fact. `journal-de-paris-1789-vol1.toml` fuses them, which is why it carries a prose
-comment explaining that its page numbers are off by one. That comment is a schema gap.
-
-**Inherited data is copied by hand.** Every issue file repeats its parent's licence,
-attribution, and source PDF. Every Turgot sheet file repeats the author, date,
-licence, and attribution of the atlas. `author = "Louis Bretez (survey and drawing);
-Claude Lucas (engraving); Aubin (lettering)"` is a string no program can parse,
-duplicated 21 times.
-
-## Settled decisions
-
-These were decided during design and are not open:
-
-1. **Four layers**: `source`, `copy`, `document`, `page`.
-2. **Flat ids.** Ids never contain `/`. `journal-de-paris-1789-vol1`, not a path.
-3. **One file per document.** File grain stays as it is today; the fix for
-   repetition is inheritance, not consolidation.
-4. **Pages are not entities.** They get no ids of their own, no special naming
-   scheme, and no files of their own. A page is declared inline as `[[page]]` and
-   addressed as `<id>.p<n>`.
-5. **`.pN` is document-local.** `n` is scoped to the document that declares it.
-6. **EDTF dates** (ISO 8601-2:2019), always quoted strings, always Gregorian.
-7. **Every page states both its numbers**; counting is the default; the validator
-   errors when the numbers do not fit.
-8. **All sources live under `source/`.**
-
-Rejected along the way: text-span and named-region citation grain (too much
-bookkeeping for the benefit); consolidating many documents into one file per volume;
-a flat relational layout; Republican-calendar machinery.
-
-## The model
-
-### Four layers
-
-Named after TEI equivalents where TEI has a good word.
-
-| `layer` | TEI analogue | what it is | example id |
-|---|---|---|---|
-| `source` | `teiCorpus` / `sourceDesc` | the publication or work, independent of any copy | `journal-de-paris` |
-| `copy` | `witness` / `msDesc` | one physical object that was digitized | `journal-de-paris-1789-vol1` |
-| `document` | `msItem` | the citable intellectual unit | `journal-de-paris-1789-01-03` |
-| `page` | `surface` | one scanned side, carrying graphics | addressed `…-01-03.p1` |
-
-`graphic` (TEI `<graphic>`) is an image file. `zone` (TEI `<zone>`) is a rectangle on
-a page, available for crops but not required.
-
-Two words were deliberately retired. **`collection`** is already taken in TEI, where
-`<collection>` means the named collection *within a holding repository* ("the David
-Rumsey Map Collection") — using it for the publication burns the word needed to record
-where a copy is held. **`issue`** is a serials word that cannot honestly describe a map
-sheet, a play, or a letter; it is one `type` of `document`, not the category itself.
-
-### Two structural fields replace `kind`
-
-- **`layer`** — which of the four. Closed set, structural, always stated.
-- **`type`** — genre within the layer. `newspaper`/`map`/`diary` on a source;
-  `issue`/`supplement`/`sheet`/`letter`/`play`/`engraving` on a document.
-
-`type = "issue"` is no longer ambiguous because `layer = "document"` has already said
-what kind of thing is being typed.
-
-### Layers may be skipped
-
-`of` points at *any* ancestor, not necessarily the layer directly above. When a layer
-is skipped, its fields collapse into the nearest declared ancestor.
-
-This is load-bearing, not a convenience. Most sources in this archive are one-off
-engravings and pamphlets, and they must not pay four files of ceremony. Journal de
-Paris genuinely needs the `copy` layer — two volumes, two shelfmarks, two PDFs, 181
-issues sharing one file. Turgot does not: one atlas, one digitization, so the Rumsey
-information sits on the source and pages address as `turgot-1739.p0`.
-
-The four layers are a vocabulary, not a required depth.
-
-## Identity and addressing
-
-- Every `source`, `copy`, and `document` has an `id`.
-- `id` defaults to the filename stem and must be globally unique across the repo.
-- Ids never contain `/`; `:` and `.` are reserved as separators.
-- Pages have no `id`. A page is `<owner-id>.p<n>` where `n` is the page's `n` field
-  within the document (or other owner) that declares it.
-
-```
-journal-de-paris-1789-01-03        a document
-journal-de-paris-1789-01-03.p1     its first page
-turgot-1739.p0                     the atlas key sheet
-```
-
-Because ids are flat and location-independent, **directories are cosmetic**. Folders
-may be reorganized at any time without breaking a single citation in the consuming
-repository. The `1789/01/03/` tree exists because it is pleasant to browse.
-
-The per-sheet `turgot_NN.toml` and `verniquet_NN.toml` files are folded into their
-source files and deleted, so underscores survive only on the `.jp2` filenames — where
-they are harmless, since a graphic is named explicitly by `graphic.file` and nothing
-parses it. The 94 image files are not renamed; they are in LFS and their names carry no
-meaning.
-
-## Repository layout
-
-```
-source/
-  journal-de-paris/
-    journal-de-paris.toml                     source
-    1789/
-      journal-de-paris-1789-vol1.pdf
-      journal-de-paris-1789-vol1.toml         copy
-      journal-de-paris-1789-vol2.pdf
-      journal-de-paris-1789-vol2.toml         copy
-      01/03/
-        journal-de-paris-1789-01-03.toml            document
-        journal-de-paris-1789-01-03-supplement.toml document
-  turgot/
-    turgot-1739.toml            source; copy collapsed in, 21 pages inline
-    turgot_00.jp2 … turgot_20.jp2
-  verniquet/
-    verniquet-1795.toml         source; copy collapsed in, 73 pages inline
-    verniquet_00.jp2 … verniquet_72.jp2
-  engravings/
-docs/
-```
-
-## File format
-
-### Source
-
-```toml
-id        = "journal-de-paris"
-layer     = "source"
-type      = "newspaper"
-title     = "Journal de Paris"
-language  = "fr"
-place     = "Paris"
-founded   = "1777-01-01"
-frequency = "daily"
-note      = "First daily newspaper published in France."
-
-[rights]
-work = "PD-old-100-expired"
-
-[[link]]
-rel = "index"
-url = "https://gazetier-revolutionnaire.gazettes18e.fr/periodique/journal-de-paris-1799"
-```
-
-### Copy
-
-```toml
-id     = "journal-de-paris-1789-vol1"
-layer  = "copy"
-of     = "journal-de-paris"
-title  = "Journal de Paris, année 1789, volume 1 (janvier–juin)"
-covers = "1789-01-01/1789-06-30"
-
-[scan]
-file  = "journal-de-paris-1789-vol1.pdf"
-count = 888
-by    = "Google Books"
-url   = "http://books.google.fr/books?id=wjkTAAAAQAAJ"
-note  = "Google front-matter leaf removed; graphic page indices are 1-based into the trimmed PDF."
-
-[holding]
-repository = "Bibliothèque cantonale et universitaire, Lausanne"
-shelfmark  = "1094184846"
-
-[identifier]
-google_books = "wjkTAAAAQAAJ"
-
-[rights]
-work        = "PD-old-100-expired"
-attribution = "Digitised by Google Books."
-```
-
-### Document
-
-Terse form, which is what 99% of issues will use:
-
-```toml
-id    = "journal-de-paris-1789-01-03"
-layer = "document"
-of    = "journal-de-paris-1789-vol1"
-type  = "issue"
-no    = 3
-date  = "1789-01-03"
-pages = { from = 13, to = 16 }
-```
-
-`pages = { from, to }` expands by counting into four pages: `n` = 1…4, graphic page =
-13…16, graphic file inherited from the copy's `scan.file`. Nothing is repeated from the
-parent.
-
-Supplements point at an id rather than a bare integer:
-
-```toml
-id    = "journal-de-paris-1789-01-03-supplement"
-layer = "document"
-of    = "journal-de-paris-1789-vol1"
-type  = "supplement"
-supplement_to = "journal-de-paris-1789-01-03"
-date  = "1789-01-03"
-pages = { from = 17, to = 20 }
-```
-
-A one-off with no parent at all — this is the shape most of the archive will take:
-
-```toml
-id    = "serment-du-jeu-de-paume-1791"
-layer = "document"
-type  = "engraving"
-title = "Le Serment du Jeu de Paume"
-date  = "1791"
-
-[[resp]]
-name = "Jacques-Louis David"
-role = "artist"
-
-[[page]]
-n = 1
-[[page.graphic]]
-file   = "serment-du-jeu-de-paume.jpg"
-width  = 4000
-height = 2800
-
-[rights]
-work = "PD-old-100-expired"
-```
-
-### Page
-
-**Pages are always declared inline**, as `[[page]]` within the file that owns them.
-There is no such thing as a page file. `page` remains one of the four layers, but it is
-the one layer that is never a file of its own — which means `layer` on a file is only
-ever `source`, `copy`, or `document`.
-
-Map sheets are the case that tempts you the other way, and they are exactly where inline
-pays off: Turgot's 21 sheets differ only in `n`, `title`, and their graphic, so as a
-single array they read as a table and diff as a table. Twenty-one files, each repeating
-`of` and `layer`, would show the same information as twenty-one separate diffs.
-
-```toml
-# source/turgot/turgot-1739.toml — continues the source file shown above
-
-[[page]]
-n     = 0
-title = "key sheet"
-[[page.graphic]]
-file   = "turgot_00.jp2"
-width  = 23964
-height = 16934
-url    = "https://www.davidrumsey.com/rumsey/download.pl?image=/166/10059022.jp2"
-
-[[page]]
-n = 1
-[[page.graphic]]
-file   = "turgot_01.jp2"
-width  = 23964
-height = 16934
-```
-
-Addresses are `turgot-1739.p0`, `turgot-1739.p1`, and so on. Turgot states `n = 0`
-explicitly because the atlas numbers its own sheets from zero; the counting default
-starts at 1, and any owner that numbers itself differently says so on its first page.
-
-### Responsibility
-
-`[[resp]]` replaces the unparseable author string, stated once on the source and
-inherited by all 21 sheets:
-
-```toml
-[[resp]]
-name = "Louis Bretez"
-role = ["surveyor", "draughtsman"]
-[[resp]]
-name = "Claude Lucas"
-role = "engraver"
-[[resp]]
-name = "Aubin"
-role = "lettering"
-```
-
-### Dates
-
-Every date field is a **quoted EDTF string**, never a TOML native date. Unquoted `1739`
-is a TOML integer and unquoted `1789-01-03` is a TOML local date, but `1789?`, `178X`,
-and `1791/1799` are neither — mixing types by how certain a date happens to be would be
-miserable to consume.
-
-Dates are always Gregorian. Where a document prints a Republican date, it is converted;
-if the conversion is uncertain, EDTF says so.
-
-| EDTF | meaning |
-|---|---|
-| `"1789-01-03"` | that day |
-| `"1739"` | that year |
-| `"1791/1799"` | interval |
-| `"1795~"` | approximate |
-| `"1789-01?"` | uncertain |
-| `"178X"` | unspecified digit |
-| `"../1789-06-30"` | open start |
-
-### Links and identifiers
-
-`url` is the canonical landing page. `[[link]]` with a `rel` carries everything else,
-retiring the ad-hoc `fetch` and `index` fields. `[identifier]` holds catalogue
-identifiers, retiring `google_books_id`.
-
-### Text and OCR
-
-Page-grain, matching the citation grain, and never inline in TOML:
-
-```toml
-[[text]]
-file = "journal-de-paris-1789-01-03.p1.txt"
-kind = "ocr"          # or "transcription"
-by   = "tesseract 5.3 fra"
-lang = "fr"
-```
-
-`by` exists because the archive will be re-OCR'd eventually, and the pages still on a
-bad pass must be identifiable.
-
-## Inheritance
-
-A file states only what is new or different. A resolver walks the `of` chain and fills
-in the rest.
-
-Inheritance is an **allowlist**, not a denylist:
-
-| inherits | never inherits |
-|---|---|
-| `[rights]`, `[holding]`, `[identifier]` | `id`, `of`, `layer`, `type`, `title`, `short_title` |
-| `[[resp]]`, `language`, `place` | `date`, `n`, `no`, `note` |
-| `scan.by`, `scan.url`, `scan.file` | `scan.count`, `scan.note`, `pages`, `covers` |
-| | `url`, `[[link]]`, `[[page]]`, `[[text]]` |
-
-`scan.file` inherits so that a document's terse `pages = { from, to }` form knows which
-PDF it is indexing into. `url` and `[[link]]` do not inherit: every layer has its own
-landing page, and silently borrowing a parent's would misattribute it.
-
-An allowlist because the failure mode matters. An over-inherited `date` would silently
-stamp every issue with `1789` and nothing would error.
-
-Merge semantics:
-
-- **Scalars** — child wins.
-- **Tables** — merged key by key; child's key wins.
-- **Arrays** (`[[resp]]`, `[[graphic]]`) — child **replaces** wholesale
-  rather than appending. Replacement is chosen so a wrong inherited value is always
-  correctable; restate the entries worth keeping, or write `resp = []` to clear.
-
-## Pages and graphics
-
-Every page states two numbers, and there is no arithmetic anywhere:
-
-- `page.n` — what page it is, within its document. This is what `.pN` matches.
-- `graphic.page` — what page it is within the graphic file. Omitted when the graphic
-  is a standalone image rather than a multi-page container.
-
-```toml
-[[page]]
-n = 1
-[[page.graphic]]
-file = "journal-de-paris-1789-vol1.pdf"
-page = 13
-```
-
-**If not specified, assume counting.** `pages = { from = 13, to = 16 }` expands to
-`n` = 1,2,3,4 against graphic pages 13,14,15,16. A page that omits `n` continues the
-count from the previous page.
-
-**The validator errors when the counting does not fit** — a range whose length
-disagrees with an explicit page list, a graphic page beyond the copy's `scan.count`, a
-duplicated `n`. Counting is a convenience, never a guess that is allowed to stand.
-
-## Validation
-
-The linter is load-bearing here, because document-local `.pN` addressing was chosen
-knowing it gives up eyeball-verifiable citations. These checks buy that back:
-
-1. Ids globally unique; every `of` resolves; no cycles in the `of` chain.
-2. Child `layer` at or below parent's, in the order `source` > `copy` > `document` >
-   `page`.
-3. `n` unique within its owner.
-4. **Sibling documents' `pages` ranges do not overlap** within a copy.
-5. Every `pages` range fits inside the copy's `scan.count`.
-6. Every `date` parses as EDTF, and a document's date falls inside its copy's `covers`
-   interval.
-7. Every `graphic.file` exists on disk, and `width`/`height` match the real file.
-8. Cross-references (`supplement_to`, and similar) resolve to existing ids.
-9. Counting expansions fit, per the rule above.
-10. Gap report: missing issue numbers in a serial's expected run.
-
-Checks 4 and 6 catch the error class that is no longer visible by eye — an issue whose
-pages silently overlap its neighbour, or one filed under the wrong volume.
-
-## Migration
-
-The existing 99 TOML files are mechanically convertible; nothing needs re-research.
-
-An earlier draft of this spec claimed 516, on the mistaken belief that the Journal de Paris
-daily issues had been transcribed. They have not. The archive holds 3 Journal de Paris
-records (the serial and its two volumes), 22 Turgot (atlas + 21 sheets) and 74 Verniquet
-(atlas + 73 sheets). The `1789/MM/DD/` directories exist but are empty — a walk of the
-directory tree looks populated, which is where the wrong number came from. Entering the
-issues is data work still to be done, and the schema is built to receive it.
-
-1. Move everything under `source/`.
-2. Add `layer` to every file; split the current `kind` value into `layer` + `type`.
-3. Hoist `licence`, `attribution`, `author`, and `url` out of leaf files into the
-   nearest ancestor that owns them; delete the copies.
-4. Convert `author` strings to `[[resp]]` records. This is the one step needing
-   judgment, and there are only a handful of distinct strings.
-5. Convert `date` values to quoted EDTF.
-6. Rename `source = "…vol1.pdf"` on documents to inherited `scan.file` on the copy.
-7. Convert `pages = "13-16"` to `pages = { from = 13, to = 16 }`.
-8. Fold the 21 `turgot_NN.toml` and 73 `verniquet_NN.toml` files into `[[page]]`
-   arrays in `turgot-1739.toml` and `verniquet-1795.toml`, then delete them. Each
-   contributes `n`, `title`, and one `[[page.graphic]]` carrying `file`, `width`,
-   `height`, and `url`; their repeated `author`, `date`, `licence`, `attribution`,
-   and `of` are dropped as inherited. The `.jp2` files are not renamed.
-9. Run the validator; fix what it finds.
-
-## Deliberately excluded
-
-- **Zones/crops as first-class citable entities.** A crop is a page plus a rectangle;
-  the page is the stable address and the rectangle is the consumer's business. `zone`
-  is available but nothing requires it.
-- **Structured facts, entities, and translations.** These live in the consuming
-  repository and cite into this one.
-- **Text spans.** Citation grain is the page.
-- **Generated titles.** Titles stay hand-written strings for now; a `title_pattern`
-  on the parent is a possible later optimization, not part of this design.
-- **Standalone page files.** An earlier draft let a page be either inline or its own
-  file. With Turgot and Verniquet inline, nothing in the archive used the file form,
-  so it was removed rather than left as an unused second code path. Reinstating it
-  later costs one loader branch.
-
-## Open question for review
-
-"All sources go in a source subfolder" is implemented above as a single top-level
-`source/` directory containing `journal-de-paris/`, `turgot/`, `verniquet/`, and
-`engravings/`. If the intent was instead that each source keeps its own folder at the
-repo root (already true today), that changes step 1 of the migration only.

docs/superpowers/specs/2026-08-07-frc-corpus-ingest-and-ocr-format-design.md

@@ -1,289 +0,0 @@
-# The Newberry French Revolution corpus — ingest, and a TOML format for OCR
-
-Status: implemented
-Date: 2026-08-07
-Revised: 2026-08-07, after the design met the corpus. The corrections are marked
-**Corrected** and the original claim is stated alongside, because most of them were
-wrong in a way worth remembering.
-
-## Purpose
-
-Bring the Newberry Library's French Revolution Collection into this archive: 38,377
-Internet Archive items of French pamphlets, mostly 1789–1799, listed by
-[NewberryDIS/frc-data](https://github.com/NewberryDIS/frc-data). For each item, hold the
-scanned PDF and the word-level OCR, addressed the same way as everything else here.
-
-That is a factor of roughly seven thousand more records than the archive currently holds.
-The design is shaped almost entirely by that number.
-
-## What frc-data actually is
-
-**Corrected.** The first draft of this said frc-data held no XML OCR, and that the
-coordinate-bearing OCR would have to be fetched from archive.org. That was wrong, and it
-was wrong because GitHub's tree API silently truncates: it returned 72,499 entries with
-`"truncated": true`, and the counts happened to add up so neatly against two directories
-that the third was never missed. `git ls-tree` on a blobless clone gives the real list.
-
-There are four directories, and each of the three that matter holds exactly 38,377 files:
-
-| directory | contents |
-| --- | --- |
-| `Metadata/*_meta.xml` | 38,377 Internet Archive metadata records |
-| `XML_for_OCR/*_djvu.xml` | 38,377 DjVu XML files — **word-level OCR with coordinates** |
-| `OCR_text/*_djvu.txt` | 38,377 plain-text OCR files |
-| `IA_pamphlets-1560-1660-marc_records/` | one 3.5 MB MARC file for a different collection |
-
-So frc-data supplies the item list, the catalogue metadata **and the OCR**. Only the PDFs
-have to be fetched.
-
-That changes the crawl from about 115,000 requests to about 77,000, and saves archive.org
-11 GB of egress for bytes that are already sitting in a git repository. Given that "be
-respectful of rate limits" was the constraint, this is not a small correction.
-
-## Scale, measured
-
-Sampled across three items and extrapolated by OCR-text size, which tracks page count
-closely:
-
-| | |
-| --- | --- |
-| items | 38,377, all with OCR |
-| PDFs | ~63 GB, still an estimate — the fetch has not finished |
-| `XML_for_OCR` in frc-data | 11 GB **measured** |
-| the same OCR as markdown | ~1.2 GB — text only, no coordinates |
-| HTTP requests | ~77,000, all of them for PDFs |
-| wall time at one request per second | ~21 hours |
-
-The wall time is the governing constraint. This is a pipeline that runs for days and is
-killed and resumed many times, not a script that is run once.
-
-## Decisions
-
-### One IA item becomes one `source` record
-
-`layer = "source"`, with `copy` and `document` collapsed in, exactly as `pellet-1873` and
-`turgot-1739` already do. A pamphlet is a single work, digitised once, and paying three
-files of ceremony for it would produce 115,000 files to describe 38,377 pamphlets.
-
-`type` comes from `mediatype` and `physical_description`: `pamphlet` under ~50 pages,
-`book` above, `periodical` where the metadata says so.
-
-### Ids are the bare Internet Archive identifier
-
-`abondance00unse`, not `frc-abondance00unse`. The identifier is already globally unique
-and stable, and it is what every other system cites. Prefixing would add nine characters
-to 38,377 addresses to solve a collision problem that does not exist — the existing
-hand-curated ids are all of the form `name-year`, which no IA identifier matches.
-
-### Directories are sharded two deep by identifier prefix
-
-```
-sources/frc/ab/abondance00unse/
-                 abondance00unse.toml
-                 abondance00unse.pdf
-                 abondance00unse.ocr.toml
-```
-
-~700 shard directories of ~55 items each. The README's promise that "directories are
-cosmetic" is what makes this free: no address changes if the sharding is ever redone.
-
-A single flat directory of 38,377 entries was rejected — not because git minds, but
-because `ls`, tab completion, and the editor's file tree all become unusable.
-
-### The OCR sidecar attaches through the existing `[[text]]` table
-
-No new top-level table. `[[text]]` already means "OCR or transcription, never inline",
-which is exactly what this is:
-
-```toml
-[[text]]
-file = "abondance00unse.ocr.toml"
-kind = "ocr"
-by   = "ABBYY FineReader 11.0"
-lang = "fr"
-```
-
-## The OCR format
-
-**Superseded twice.** It was TOML carrying packed word-level arrays when this was written; it
-is now markdown with YAML frontmatter carrying no coordinates at all. Both changes were at the
-repository owner's direction, and the second is the larger one.
-
-### Where it landed
-
-One file per scanned page, `<id>.pN.ocr.md`:
-
-```markdown
----
-of: 1789iemilseptcen00unse
-page: 6
-engine: ABBYY FineReader 11.0
-lang: fr
-w: 2597
-h: 4418
-dpi: 500
----
-
-<the page's text>
-```
-
-Frontmatter says which page of what and how big the image is; the body is the page. That is
-the whole format.
-
-### What was dropped, and what that costs
-
-Gone: word boxes, per-word confidence, baselines, region background colours, and the `brk`
-encoding that reconstructed DjVu's four levels of nesting in one byte per word. With them goes
-the losslessness claim and `scans ocr export`, which existed to make that claim executable.
-
-The reason is proportion. The coordinates were about seven eighths of the bytes — 8.3 GB
-against roughly 1.2 GB for the text — and nothing was reading them.
-
-**Nothing is destroyed.** The coordinates are in `XML_for_OCR/` in frc-data, which is one
-clone away and needs no request to archive.org. Reinstating them is a change to
-`tools/src/djvu.rs` and a re-run of `frc ingest`, not a re-fetch.
-
-### The per-page split
-
-Before the coordinates were dropped, one file per *item* put a 3,378-page bound run of nine
-volumes into a single 49 MB text file. Splitting per page fixed that, and kept two properties
-worth naming even now that files are small:
-
-- `<id>.pN` matches the address grammar the archive already uses for pages.
-- The body needs no `## Page N` headings, so nothing in the OCR can be mistaken for a
-  delimiter. The per-item version had to carve the body back up by markers the OCR could in
-  principle contain.
-
-### Bare scalars, and the YAML version that matters
-
-Values are written bare — `lang: fr` — because the frontmatter is meant to read as YAML.
-
-The hazard is that `no` is the ISO 639-1 code for Norwegian and, in **YAML 1.1**, the boolean
-false. `serde_norway` implements **1.2**, where it is already a string, so a round-trip check
-against this crate's own parser passes it. That check is not the right bar: PyYAML, libyaml,
-Psych and gopkg.in/yaml.v2 are all 1.1, and this data is published for their tools rather than
-ours. So there is an explicit 1.1 deny-list *and* a round-trip check, and a test asserts that
-every reserved word survives as the string it was.
-
-## New record fields
-
-**Corrected.** The draft proposed a fourth, an `[ocr]` table pointing at the sidecar. It
-was not needed: `[[text]]` already means "OCR or transcription, never inline", which is
-exactly what a sidecar is. Three remain, each justified independently of this ingest:
-
-| field | why |
-| --- | --- |
-| `subject` — array of strings | 38,377 pamphlets are unusable without subject access. LCSH terms, from IA's `<subject>`. |
-| `extent` — string | The bibliographic collation, `"7 p. ; 18 cm."`, verbatim. Distinct from `scan.count`, which counts scanned images including covers. |
-| `scan.ppi` — integer | The scan's own resolution. `render` currently infers native resolution from the largest embedded image; this states it. |
-
-`subject` and `extent` go on all three record variants and do not inherit. `ppi`
-inherits, like the rest of `[scan]` except `count` and `note`.
-
-## Metadata mapping
-
-| IA field | record |
-| --- | --- |
-| `title` | `title` |
-| `date`, `dfate` (a real IA typo, present in some records) | `date`, as EDTF |
-| `creator`* | `[[resp]]`; a trailing `, printer` inside the name becomes `role` |
-| `publisher` | `place` (before the first ` : `) and a `[[resp]]` with `role = "publisher"` |
-| `language` | `language`, ISO-639-2 → BCP-47 (`fre` → `fr`) |
-| `contributor`, `call_number`, `collection` | `[holding]` repository, shelfmark, collection |
-| `identifier`, `identifier-ark`, `openlibrary_edition`, `openlibrary_work` | `[identifier]` |
-| `link_to_catalog`, `identifier-access` | `[[link]]` with `rel` `catalogue`, `viewer` |
-| `imagecount`, `ppi`, `scanner`, `camera`, `scandate`, `scanningcenter` | `[scan]` |
-| `subject`* | `subject` |
-| `physical_description` | `extent` |
-| `description`*, `notes`, `citation` | `note` |
-| `sponsor` | `rights.attribution` |
-
-Rights are constant across the corpus: the works are of the 1790s and long out of
-copyright; the digitisation is the Newberry's, sponsored and public.
-
-## The pipeline
-
-A new `scans fetch` subcommand, behind a `fetch` feature so the default build stays free
-of an HTTP stack.
-
-**Resumability** is the first requirement, not an afterthought. **Corrected:** the draft
-specified a JSONL ledger. There is none. An item is done when its file is on disk, which is
-the whole resume condition — a ledger would be a second source of truth that can fall out
-of step with the tree it describes, bought at the price of a file to keep consistent.
-Downloads are written to `.pdf.part` and renamed, so a run killed mid-download cannot leave
-a half PDF that the next run mistakes for a finished one.
-
-**Politeness** is the second:
-
-- one request in flight at a time, minimum interval configurable, default 1s
-- a `User-Agent` naming the project and its repository, so the operator can be contacted
-- `Retry-After` honoured on 429 and 503
-- exponential backoff on 5xx, capped, then the item is parked rather than blocking the queue
-- `--limit` and `--since` so a run can be bounded
-
-**Verification**: every download is checked for size and MD5 against the values in the
-item's IA metadata before it is accepted. A truncated PDF must not be committed and then
-believed.
-
-**Disposal**: `_djvu.xml` is converted to `.ocr.toml` and deleted. Keeping both would
-double 12 GB to hold the same facts twice, and the TOML is the archival form.
-
-### Storage
-
-| what | where | why |
-| --- | --- | --- |
-| `*.pdf` | Git LFS | ~63 GB, already compressed, no benefit from packfiles |
-| `*.ocr.toml` | plain git | text, packs about 5:1; LFS would bill for the full 7 GB |
-| `*.toml` records | plain git | small |
-
-63 GB of LFS exceeds GitHub's free tier and needs data packs on the remote before the
-first push. Pushing proceeds in commits of a few hundred items rather than one enormous
-commit, so an interrupted push loses minutes rather than days.
-
-## Validation
-
-Check 11, four codes:
-
-- **E708** a `[[text]]` pointing at a sidecar that is not there
-- **E709** a sidecar that is not valid TOML for the OCR model
-- **E710** a sidecar whose `of` names a record other than the one pointing at it
-- **E711** any of the internal inconsistencies in `ocr::Problem` — arrays of unequal length,
-  `box` not exactly four per word, `region-bg` not one per region, `text` that disagrees
-  with `word` and `brk`
-
-**Corrected.** The draft said these "run over the whole archive by default, since they read
-only TOML". That reasoning was wrong: the distinction is not TOML versus bytes, it is how
-many gigabytes of them there are. Parsing 8 GB of sidecars turns `validate` from a
-2.3-second command into a 2m37s one. E708 is a `stat` and runs always; everything that has
-to parse a sidecar is behind `--probe`, for exactly the reason the image checks are.
-
-Result on the full corpus: 38,715 records, **0 errors, 0 warnings** by default, and 0 errors
-with `--probe` across all 38,377 sidecars.
-
-One further correction fell out of the first full run: `id_is_preferred` treated `_` as
-outside house style, which made 1,263 warnings out of identifiers like `procesverbal00_1_0`
-and `case_oversize_frc_27598`. Those are the handles archive.org publishes and everyone else
-cites. Underscore is now a group separator alongside hyphen — 1,263 warnings a reader cannot
-act on is how a warning channel stops being read.
-
-## Testing
-
-- djvu.xml → TOML → djvu.xml round-trip over a sample of real fetched items
-- the `brk` encoder and decoder, against hand-built nestings including the degenerate
-  cases: an empty page, a page of one word, a word containing a space
-- metadata mapping over a fixture set chosen for awkwardness: missing `date`, the `dfate`
-  typo, multiple `creator` entries, a creator with an embedded role, no `publisher`
-- ledger resume: interrupt mid-run, restart, confirm no item is fetched twice and none is
-  skipped
-- rate limiter timing, with a clock injected rather than by sleeping
-
-## Deliberately not done
-
-- **No `copy` or `document` layer per item.** Revisit only if an item turns out to bind
-  several distinct works, which the metadata does not currently distinguish.
-- **No text search index.** Grep over 7 GB is adequate; a real index is a different
-  project with a different lifetime.
-- **No re-OCR.** ABBYY 11 on 500 ppi scans of worn 18th-century type is mediocre, and
-  visibly so in the confidence figures. Improving it is worth doing and is not this.
-- **No MARC ingest.** The 3.5 MB MARC file covers a different, earlier collection
-  (1560–1660 pamphlets) than the one being ingested.

docs/superpowers/specs/2026-08-14-hardy-pdf-and-wikimedia-design.md

@@ -1,270 +0,0 @@
-# Hardy: volume PDFs and Wikimedia records
-
-2026-08-14
-
-Hardy's journal arrived from Gallica as 4,230 loose JPEGs across eight directories. This
-turns each volume into one PDF without touching a pixel, and writes the Wikidata and Commons
-metadata needed to publish them.
-
-Two products, independent of each other:
-
-* `sources/hardy/volN/hardy-volN.pdf` — one PDF per volume, carrying the Gallica JPEGs. Each
-  sits beside the record that describes it, which is where every other container in this
-  archive already lives.
-* `sources/hardy/wikimedia/` — three staged QuickStatements batches and eight Commons file
-  pages, all generated from the records rather than typed.
-
-## The PDF
-
-### Why a PDF at all
-
-The archive already holds the pages, and a PDF adds nothing a reader could not get from the
-directory. What it adds is a *single addressable object*: something to hand to Commons, to
-cite, to hash. Volume 1 is 658 files and one boundary; as a PDF it is one file and the same
-boundary, and the ordering — which leaf follows which — stops being a filename convention
-and becomes structure.
-
-The order comes from the `[[page]]` list in `hardy-volN.toml`, not from sorting the
-directory. Those are the same today. They are the same today *because* the records were
-generated in fetch order, and a record is the thing that gets corrected when a leaf turns out
-to have been shot twice or out of sequence. Reading the directory would silently ignore such
-a correction.
-
-### Native pixels, unmodified
-
-Every page is embedded as a `/DCTDecode` image XObject. Nothing is resampled, nothing is
-requantised, no colour is converted. Sampling all eight volumes finds one format throughout:
-
-```
-SOF2 progressive · 1 component · 8 bit · JFIF density 400x400 dpi
-```
-
-`MediaBox` is therefore `width/400 x height/400 x 72` points — about 725 x 1098 for a typical
-leaf, which is roughly A2. Page geometry varies leaf to leaf and the PDF preserves that
-rather than normalising to a common box: the scans genuinely differ, and flattening them
-would assert a uniformity the object does not have.
-
-`[dimensions]` on the copy record says the volume measures 355 x 235 mm. The scan at 400 dpi
-works out to about 387 x 256 mm, some nine per cent larger in both axes, which is the shot
-including a margin of surround beyond the leaf. The record's figure describes the manuscript;
-the PDF's describes the photograph. Neither is corrected against the other.
-
-### Progressive JPEG is the problem
-
-`DCTDecode` is specified against baseline JPEG. Gallica serves progressive. A progressive
-stream dropped into `DCTDecode` is off-spec, and whether it renders is a property of the
-reader rather than of the file.
-
-Measured, not assumed: poppler and Ghostscript both decode it correctly. That is most
-readers in practice, and Ghostscript in particular is what Commons thumbnails PDFs with. It
-is still off-spec, and a file that depends on the reader being lenient is not an archival
-object.
-
-So each stream is transcoded progressive to baseline before it is embedded. This is
-**lossless**: `jpegtran` entropy-decodes the scans to DCT coefficients and re-emits them
-sequentially. The coefficients are not touched, so there is no requantisation and no
-generation loss — the decoded pixels are identical, bit for bit. Only the arrangement of the
-bits describing them changes.
-
-That claim is checked rather than trusted. The builder decodes both the source JPEG and the
-stream it is about to embed, compares the raster, and fails on any difference. A silent
-change to 4,230 pages of manuscript is exactly the failure worth paying a decode for.
-
-`jpegtran` is invoked with `-copy all`, not `-copy none`. Dropping the APP segments would
-drop the JFIF density in APP0, and that density is where the page's physical size comes from;
-without it every `MediaBox` would be a guess.
-
-### Linearised, so the first page arrives first
-
-Written straight out, a PDF's cross-reference table is at the end and its pages are in
-whatever order they were emitted. A reader wanting page one of a 2.5 GB volume must therefore
-fetch the whole file. Linearising moves the first page's objects, a partial cross-reference
-table and a hint stream to the front, so a viewer speaking HTTP byte ranges renders page one
-after a few megabytes. These volumes are served from object storage and read over a network;
-without this they are technically available and practically unreadable.
-
-`qpdf --linearize --stream-data=preserve` does it. Explicitly **not** Ghostscript's
-`-dFastWebView`, which reaches linearisation through `pdfwrite` and re-encodes images on the
-way past — that would discard the losslessness the rest of this design exists to protect.
-
-The same standard applies as to the transcode: the linearizer is trusted to move objects, not
-to alter them, and that is checked rather than assumed. Every `DCTDecode` stream is read out
-of both the before and after file and compared byte for byte; one differing byte and the
-original is kept.
-
-#### …but not past 2 GB, which is every volume here
-
-This was implemented, run over all eight volumes, and then reverted. Recording why, because
-the idea is a good one and will otherwise be had again.
-
-The linearisation itself was faultless: all 4,230 image streams byte-identical afterwards,
-`/L` exactly equal to the real file length, `qpdf --check` reporting "File is linearized. No
-syntax or stream encoding errors found", and page one reachable in the first 2.6–4.2 MB of a
-2.1–2.6 GB volume — a thousandfold improvement in what a reader must fetch first.
-
-It still had to go, because a linearization table states `/L`, `/T` and its hint-stream
-offsets as byte positions into the file, and past `i32::MAX` those no longer fit the 32-bit
-integer some readers keep them in:
-
-| reader | 2.5 GB plain | 2.5 GB linearised |
-|---|---|---|
-| `hayro`, which `render` and `extract` use | every page | **1 page** |
-| poppler | clean | 21 warnings per open; recovers via the main xref |
-| Ghostscript | clean | clean |
-
-The files were not malformed — Ghostscript and qpdf both accept them, and the same qpdf
-invocation on a 12.8 MB slice of the same volume warns from nobody, which is what identifies
-size rather than output quality as the cause. But breaking the decoder the archive's own tools
-depend on, to speed up a first-page load for the subset of readers that handle it, is a bad
-trade. `scans pdf` now skips linearisation above `i32::MAX` bytes and says why.
-
-The way to have both is to get under 2 GB, which means splitting volumes — considered for
-Commons and declined there, and declined here for the same reason: it costs the
-one-file-per-volume correspondence between the archive and the object.
-
-### jpegtran and qpdf are operator tools, not dependencies
-
-The crate stays pure Rust with no C toolchain, which is a constraint the README makes a
-point of. `mozjpeg-sys` would satisfy the transcode and break that.
-
-Both are instead invoked as external programs, in the same category as `git-lfs` and a
-network connection: something the operator has when running an ingest, not something the
-build links against. `scans pdf` checks for both up front and says what to install if either
-is missing, rather than failing 400 pages in.
-
-### Command
-
-```sh
-cargo run --release --features pdf -- pdf hardy-vol1
-cargo run --release --features pdf -- pdf hardy          # all eight
-```
-
-Behind a `pdf` feature, off by default — but for a narrower reason than the other gated
-features. Writing the PDF needs no dependency at all: it is bytes, and this crate writes them
-by hand. *Verifying* the transcode needs a JPEG decoder, and `validate` must not carry one.
-
-The decoder is `zune-jpeg`, deliberately not the `djpeg` that ships beside `jpegtran`. The
-transcode is performed by libjpeg, so decoding both sides with libjpeg would establish only
-that libjpeg agrees with itself. An independent pure-Rust implementation reading both files
-is an actual second opinion.
-
-The argument is an address, like every other command here. `hardy` resolves to the source and
-builds each copy under it; `hardy-vol1` builds one.
-
-## The Wikimedia records
-
-Nothing here talks to a network. The output is files to be reviewed and then pasted, so that
-every edit that lands on a public project has been read by a person first.
-
-### Wikidata: nine items
-
-Wikidata has no item for this work, and Hardy (Q3484827) has no linked works at all.
-
-The structure mirrors the archive's own `source` / `copy` split, because the same fact
-pattern forces it: each volume has its own shelfmark, its own Gallica ARK, its own date range
-and its own folio count. Those are statements about eight different subjects, and modelling
-them as qualifiers on one item would make them unqueryable.
-
-```
-work    Mes loisirs, ou Journal d'evenemens tels qu'ils parviennent a ma connoissance
-        P31 Q185598 diary · P50 Q3484827 · P407 Q150 · P495 Q142 · P1476 title
-        P217 Francais 6680-6687 · P195 Q193563 · P6216 Q19652 · P2635 8
-        P580/P582 from `covers` · P973 Gallica landing page
-
-volume  Mes loisirs, volume 1 .. volume 8
-        P31 Q87167 manuscript · P361 -> work · P50 Q3484827 · P407 Q150
-        P195 Q193563 · P217 Francais 6680..6687
-        P4258 Gallica ID · P8091 Archival Resource Key
-        P6108 IIIF manifest, from `[[link]] rel = "iiif"` — recorded, not built from the ark
-        P1104 from `[scan] count` · P580/P582 from `covers`
-        P2048/P2049 from `[dimensions]` in mm · P186 Q11472 paper
-        P6216 Q19652 public domain · P973 Gallica landing page
-```
-
-Every identifier above was checked against the live API on 2026-08-14 and is recorded in
-`wikimedia::vocab` with the label it carried, because property numbers are not memorable and
-are very easy to get subtly wrong. Two were wrong on the first pass and the check caught
-them: `P7101` is *privacy policy URL*, not Gallica ID (that is `P4258`), and `Q174728` is
-*centimetre*, not millimetre (that is `Q174789`). Each generated statement line carries the
-property's label in a trailing comment so the same class of error shows up in review as a
-wrong word rather than a wrong number.
-
-`covers` is EDTF. `1764/1771` becomes P580 1764 and P582 1771, both at year precision — the
-`/9` suffix in a QuickStatements time value. The interval is the span the volume records, not
-the span of its writing, so it is start and end time rather than inception.
-
-`P1104` is the count of **digitised views**, which includes bindings and flyleaves. It is
-what the record knows, and it is not the same as the number of written folios. The README
-says so where somebody reviewing the batch will read it.
-
-P800 is added to Hardy pointing at the work, since the item currently claims no works.
-
-**Three files, not one.** QuickStatements v1 addresses a newly created item only as `LAST` —
-"the item the previous `CREATE` made". There is no way to say "the item created eight lines
-ago", so one batch cannot both create a work and create eight volumes pointing at it. The
-output is therefore staged, `-1-work.qs`, `-2-volumes.qs`, `-3-links.qs`, with `$WORK` and
-`$VOLn` where a Q-id belongs. `--work Q…` fills the first of them in. Substituting a
-placeholder is a visible, checkable act; guessing a Q-id is not.
-
-### Commons: eight file pages
-
-One `.commons.wiki` per volume, to be pasted into the upload form.
-
-```
-File:Hardy - Mes loisirs, volume 1 (BnF Francais 6680).pdf
-```
-
-`{{Book}}` rather than `{{Information}}`: this is a bibliographic object with an author, a
-date range, a holding institution and a shelfmark, and `{{Book}}` has fields for all of them.
-`{{Institution:Bibliotheque nationale de France}}` and `{{Gallica}}` supply the source line
-Gallica's terms ask for.
-
-Every template and category referenced was checked for existence against the live Commons API
-before being emitted, which changed three things. There is no `Creator:Simeon-Prosper Hardy`
-page, so the author line is a plain name with `{{Q|Q3484827}}` beside it rather than a
-`{{Creator:…}}` that would render as a red link. `Category:Simeon-Prosper Hardy` does not
-exist either, but that one is emitted anyway — creating it is the normal order of operations
-for a new subject — with a note in the README saying so, so it does not read as a mistake.
-The editorial `note` from the record is English prose about a French manuscript, so it is
-wrapped in `{{en|…}}`: the language template declares the language of the text inside it, not
-of the object being described.
-
-**Licensing.** One tag doing two jobs: `{{PD-scan|PD-old-100-expired}}`. The argument covers
-Hardy's text — he died in 1806 — and the outer template covers the photography. Gallica
-asserts a non-commercial restriction over its digitisations, and Commons does not accept
-non-commercial licences, but it does treat faithful reproductions of two-dimensional
-public-domain works as public domain regardless of what the digitiser asserts. That is the
-settled basis on which Gallica's scans are already hosted there, and it is the position taken
-here. The repository's own `[rights]` block records Gallica's claim unchanged; it is not
-rewritten to match.
-
-The licence appears in the `== {{int:license-header}} ==` section and the `Permission` field
-is left empty. Stating it in both places invites the two copies to drift apart when one is
-edited.
-
-**File size.** Each volume lands around 2.0-2.4 GB. Commons' ceiling is 4 GB, so they fit,
-but not through UploadWizard, which stops at 100 MB — these need a chunked upload via
-pywikibot or the API. The thumbnailer may well time out on 658 pages at 4000x6000, leaving a
-file that is hosted but has no preview. Accepted knowingly: the full-resolution volume is the
-artefact worth having, and the Wikidata items stand on their own and point at Gallica's IIIF
-if Commons will not serve it.
-
-## One fix picked up on the way
-
-`[holding] wikidata` exists in the schema, and its own doc comment says it "feeds a
-`collection` claim (P195)" — but `load::ResolvedHolding` did not carry it, so it was silently
-dropped during inheritance and never reached anything that could have used it. Hardy declares
-it once on the source, where the eight volumes should inherit it.
-
-The merge is documented as `[holding]` key by key up the chain; the struct had simply
-forgotten a key. Adding `wikidata` beside `repository`, `shelfmark` and `collection` restores
-the documented behaviour rather than changing it.
-
-## What is not here
-
-* No OCR, and no text layer in the PDFs. The manuscript is cursive French hand; the OCR
-  pipeline in this repository is for print and would produce noise that reads as data.
-* No bookmarks or outline. The dated entries inside each volume are not recorded yet, and
-  inventing an outline from folio numbers would add no information.
-* No downsampled edition. It would answer the Commons thumbnail problem by re-compressing,
-  which is the one thing this design exists to avoid.

schemas/README.md

@@ -1,134 +0,0 @@
-# schemas/
-
-## source.json is GENERATED. Do not edit it.
-
-`schemas/source.json` is derived from the Rust types in `tools/src/model.rs` via
-`schemars`. It is not a hand-maintained document, and editing it by hand is the one
-thing that must never happen here: the schema would start disagreeing with the
-validator that actually enforces the archive, and a schema that lies is worse than no
-schema, because it is believed.
-
-Regenerate it after any change to the record types:
-
-```sh
-cd tools
-cargo run -- schema
-```
-
-Verify it has not drifted — this is what CI runs, and it exits non-zero when the
-checked-in file differs from what the types generate:
-
-```sh
-cd tools
-cargo run -- schema --check
-```
-
-One definition in `model.rs` drives three things at once: Rust type checking, TOML
-deserialisation, and this JSON Schema. That is the whole point of generating it.
-
-## What the schema describes
-
-A single archive record. It is a `oneOf` of three branches discriminated on `layer`:
-
-| `layer`    | required fields   |
-| ---------- | ----------------- |
-| `source`   | `layer`, `title`  |
-| `copy`     | `layer`, `title`  |
-| `document` | `layer`           |
-
-`document` deliberately does not require `title` — a terse issue record has none.
-`layer = "page"` is not a branch and is rejected: pages exist only inline, as
-`[[page]]` entries inside a record, never as files of their own.
-
-Because the discriminator is a `const` on `layer`, writing `layer = "copy"` narrows
-both autocomplete and the required-field set to the copy branch alone.
-
-## How a record gets validated in the editor
-
-Each record carries a `#:schema` directive as its first line, pointing at this file
-**relative to the record**:
-
-```toml
-#:schema ../../schemas/source.json
-layer = "source"
-title = "Plan de Turgot"
-```
-
-The number of `../` segments depends on the record's depth: two from
-`source/turgot/`, more from `source/journal-de-paris/1789/01/03/`. Whatever writes
-records — including `scans migrate` — must emit this line with the correct depth.
-
-The directive is the only portable mechanism, and that is a finding, not a preference.
-`evenBetterToml.schema.associations` cannot express a repo-relative schema path:
-
-- a relative value such as `./schemas/source.json` is resolved against an internal
-  `root:///` base rather than the workspace, so the schema fails to load;
-- `${workspaceFolder}` is not expanded;
-- a bare Windows path like `D:/…` is parsed as a URL with the scheme `d`.
-
-In all three cases the association *matches* the document and then produces **zero
-diagnostics**. Nothing is reported in the editor — the failure is visible only in the
-extension's output channel. A committed association would therefore hand every
-contributor the appearance of validation and none of the substance. `.vscode/settings.json`
-explains this at the point of temptation.
-
-Install the extension recommended in `.vscode/extensions.json`
-(`tamasfe.even-better-toml`); without it, TOML files get no schema validation at all.
-
-## Enforcing the schema outside the editor
-
-`.taplo.toml` maps `source/**/*.toml` to this schema for the
-[taplo](https://taplo.tamasfe.dev/) CLI, so the same rules apply in CI:
-
-```sh
-# validate every archive record against schemas/source.json
-taplo check
-
-# if the config's globs misbehave (seen on Windows), pass the paths explicitly
-taplo check "source/**/*.toml"
-
-# check formatting too, without rewriting anything
-taplo fmt --check --diff
-```
-
-Install with `cargo install taplo-cli --locked`.
-
-A full CI sequence, cheapest check first:
-
-```sh
-cd tools && cargo run -- schema --check   # schema matches the types
-taplo check                               # records match the schema
-cd tools && cargo run -- validate         # the 10 semantic checks
-```
-
-The schema and `scans validate` do different jobs and neither replaces the other. The
-schema checks the *shape* of one file in isolation — field names, types, which fields
-each layer requires. `scans validate` checks everything that needs more than one file
-or more than one field: that `of` resolves and does not cycle, that ids are unique,
-that page ranges fit inside `scan.count` and do not collide between siblings, that
-dates parse as EDTF and fall inside their copy's `covers`.
-
-## Verification status
-
-Confirmed by driving the bundled taplo language server directly:
-
-- `schemas/source.json` parses, is a structurally valid JSON Schema
-  (`https://json-schema.org/draft/2020-12/schema`), and its `oneOf` discriminates on
-  `layer` via three `const` branches;
-- `cargo run -- schema --check` reports the checked-in file up to date;
-- with a `#:schema` directive, an invalid record is flagged and a valid one is clean;
-  a `source` missing `title` is flagged, a `document` missing `title` is not, and
-  `layer = "page"` is rejected — the per-layer discrimination genuinely works.
-
-Not verified, and stated plainly rather than assumed:
-
-- **The `taplo` CLI itself was never run.** `cargo install taplo-cli` failed here with
-  `os error 112` — the disk holding `CARGO_HOME` is full. Every `taplo` command above
-  is written from its documented interface, not from observed output, and the
-  `.taplo.toml` rule syntax is unconfirmed against a real binary. Run `taplo check`
-  once on a machine with disk space before relying on it in CI.
-- **The archive has not been migrated yet.** The records still live in
-  `journal-de-paris/`, `turgot/` and `verniquet/` at the repo root, in the legacy
-  field layout, with no `source/` directory and no `#:schema` directives. They will
-  *not* validate against this schema, and `taplo check` currently matches no files.
-  That is expected until `scans migrate` runs; it is not a fault in the schema.

schemas/ocr.json

@@ -1,45 +0,0 @@
-{
-  "$schema": "https://json-schema.org/draft/2020-12/schema",
-  "additionalProperties": false,
-  "description": "One <id>.ocr.toml: word-level OCR losslessly equivalent to a DjVu XML derivative. Generated from the Rust types in tools/src/ocr.rs by `scans schema` - do not edit by hand.",
-  "properties": {
-    "dpi": {
-      "description": "Resolution the page was scanned at.",
-      "format": "int64",
-      "type": "integer"
-    },
-    "engine": {
-      "description": "Producing agent and version, e.g. `ABBYY FineReader 11.0`.",
-      "type": "string"
-    },
-    "h": {
-      "description": "Pixel height of the page image.",
-      "format": "int64",
-      "type": "integer"
-    },
-    "lang": {
-      "description": "BCP-47 code of the text, when known.",
-      "type": "string"
-    },
-    "of": {
-      "description": "Id of the record this page belongs to.",
-      "type": "string"
-    },
-    "page": {
-      "description": "1-based page index within the scanned container, matching `.pN` addressing.",
-      "format": "int64",
-      "type": "integer"
-    },
-    "w": {
-      "description": "Pixel width of the page image.",
-      "format": "int64",
-      "type": "integer"
-    }
-  },
-  "required": [
-    "of",
-    "page"
-  ],
-  "title": "Word-level OCR sidecar",
-  "type": "object"
-}

schemas/source.json

@@ -1,751 +0,0 @@
-{
-  "$defs": {
-    "Dimensions": {
-      "additionalProperties": false,
-      "description": "Physical size of the object, `dimensions = { height = 180, width = 115, unit = \"mm\" }`.\n\nThe object as it exists on a shelf, not the image of it — pixel size lives on\n[`Graphic`] and scanning resolution on [`Scan::ppi`]. Kept structured rather than left in\nthe `extent` string because \"18 cm.\" is a fact a catalogue can compare, sort and convert,\nand `\"7 p. ; 18 cm.\"` is not.\n\nThis is what a Wikidata `height`/`width` claim (P2048/P2049) is built from.\nWhole units, not floats. Catalogue dimensions are whole numbers — \"18 cm\", \"in-8\" — and\nmaking them integers keeps the record types comparable with `Eq` and keeps 18.0 from ever\nfailing to equal 18.000000000000004.",
-      "properties": {
-        "depth": {
-          "description": "Thickness of a bound volume, where it is known.",
-          "format": "int64",
-          "type": "integer"
-        },
-        "height": {
-          "format": "int64",
-          "type": "integer"
-        },
-        "unit": {
-          "description": "`mm` or `cm`. Always stated, because a bare number is not a measurement.",
-          "type": "string"
-        },
-        "width": {
-          "format": "int64",
-          "type": "integer"
-        }
-      },
-      "required": [
-        "unit"
-      ],
-      "type": "object"
-    },
-    "Graphic": {
-      "additionalProperties": false,
-      "description": "One image file, or one page inside a multi-page container.",
-      "properties": {
-        "file": {
-          "description": "Image or container path. Omitted means \"fall back to the resolved `scan.file`\".",
-          "type": "string"
-        },
-        "height": {
-          "description": "Pixel height of **this page**, not of the container.",
-          "format": "int64",
-          "type": "integer"
-        },
-        "mimetype": {
-          "description": "Explicit media type; otherwise inferred from the extension.",
-          "type": "string"
-        },
-        "note": {
-          "type": "string"
-        },
-        "page": {
-          "description": "1-based page index **inside** `file`. Omitted when `file` is a standalone image.",
-          "format": "int64",
-          "type": "integer"
-        },
-        "url": {
-          "description": "Direct download URL for this exact image (the retired `fetch` field).",
-          "type": "string"
-        },
-        "width": {
-          "description": "Pixel width of **this page**, not of the container.",
-          "format": "int64",
-          "type": "integer"
-        }
-      },
-      "type": "object"
-    },
-    "Holding": {
-      "additionalProperties": false,
-      "description": "Where the physical object lives. Inherits, merged key by key.",
-      "properties": {
-        "collection": {
-          "description": "The named collection within the repository, e.g. `David Rumsey Map Collection`.",
-          "type": "string"
-        },
-        "note": {
-          "type": "string"
-        },
-        "repository": {
-          "description": "Institution, e.g. `Bibliothèque cantonale et universitaire, Lausanne`.",
-          "type": "string"
-        },
-        "shelfmark": {
-          "description": "Call number. **Always a string**, never an int — leading zeros and letters occur.",
-          "type": "string"
-        },
-        "wikidata": {
-          "description": "Wikidata item for the holding institution, e.g. `Q193563` for the BnF.\n\nFeeds a `collection` claim (P195), which wants the institution's item rather than its\nname.",
-          "type": "string"
-        }
-      },
-      "type": "object"
-    },
-    "Link": {
-      "additionalProperties": false,
-      "description": "Everything that is not the canonical `url`. Retires ad-hoc `index` and `fetch`.",
-      "properties": {
-        "note": {
-          "type": "string"
-        },
-        "rel": {
-          "description": "Relation type. Open vocabulary: `index`, `catalogue`, `viewer`, `download`, `iiif`,\n`thumbnail`, `about`, `worldcat`.",
-          "type": "string"
-        },
-        "title": {
-          "type": "string"
-        },
-        "url": {
-          "type": "string"
-        }
-      },
-      "required": [
-        "rel",
-        "url"
-      ],
-      "type": "object"
-    },
-    "Page": {
-      "additionalProperties": false,
-      "description": "One scanned side, carrying graphics. **Declared inline only** — there is never a page file.",
-      "properties": {
-        "dimensions": {
-          "$ref": "#/$defs/Dimensions",
-          "description": "Physical size of this leaf, where it differs from the object as a whole — an atlas\nsheet or a folding plate."
-        },
-        "graphic": {
-          "description": "Zero or more. Zero is legal (a text-only page) and emits `W904`.\nThe **first** entry is the primary graphic: the one `.pN` resolves to and the one\nwhose pixel space zone coordinates are in.",
-          "items": {
-            "$ref": "#/$defs/Graphic"
-          },
-          "type": "array"
-        },
-        "id": {
-          "description": "A stable id for this page in its own right, e.g. `hardy-v1-p027`, `turgot-sheet01`.\n\nOptional, and worth setting when the page is an artefact somebody would describe on\nits own: an atlas sheet, a plate, a folio of a manuscript. It is an *addition* to\n`<owner>.pN`, not a replacement — both resolve to the same page — and it exists so a\nWikidata item or a Commons upload can cite the sheet rather than \"page 4 of the\nthing that contains it\".\n\nMust be unique across the whole archive, and must not collide with a record id.",
-          "type": "string"
-        },
-        "label": {
-          "description": "The printed page label when it differs from `n`, e.g. roman `xxxvj`.\nPurely descriptive: never parsed, never used for addressing.",
-          "type": "string"
-        },
-        "link": {
-          "items": {
-            "$ref": "#/$defs/Link"
-          },
-          "type": "array"
-        },
-        "n": {
-          "description": "Page number **within its owning document**. This is what `.pN` matches.\nOmitted means \"continue the count\". May be `0` (Turgot) or negative.",
-          "format": "int64",
-          "type": "integer"
-        },
-        "note": {
-          "type": "string"
-        },
-        "text": {
-          "items": {
-            "$ref": "#/$defs/Text"
-          },
-          "type": "array"
-        },
-        "title": {
-          "type": "string"
-        },
-        "url": {
-          "description": "The page's own landing page.",
-          "type": "string"
-        },
-        "zone": {
-          "items": {
-            "$ref": "#/$defs/Zone"
-          },
-          "type": "array"
-        }
-      },
-      "type": "object"
-    },
-    "PageRange": {
-      "additionalProperties": false,
-      "description": "A terse page range, `pages = { from = 13, to = 16 }`.",
-      "properties": {
-        "from": {
-          "format": "int64",
-          "type": "integer"
-        },
-        "to": {
-          "format": "int64",
-          "type": "integer"
-        }
-      },
-      "required": [
-        "from",
-        "to"
-      ],
-      "type": "object"
-    },
-    "Resp": {
-      "additionalProperties": false,
-      "description": "A statement of responsibility. Replaces the unparseable `author` string.",
-      "properties": {
-        "name": {
-          "description": "Person or corporate name as written.",
-          "type": "string"
-        },
-        "note": {
-          "description": "Qualifications such as \"attributed\".",
-          "type": "string"
-        },
-        "role": {
-          "$ref": "#/$defs/Roles",
-          "description": "One role or several. Both spellings are legal."
-        },
-        "wikidata": {
-          "description": "Wikidata item for this person or body, e.g. `Q3484827`.\n\nA name string is not an identity — there are three 18th-century Parisian booksellers\ncalled Hardy — and a Wikidata `author` claim (P50) needs the item, not the string.",
-          "type": "string"
-        }
-      },
-      "required": [
-        "name"
-      ],
-      "type": "object"
-    },
-    "Rights": {
-      "additionalProperties": false,
-      "description": "Rights status. Inherits, merged key by key.",
-      "properties": {
-        "attribution": {
-          "description": "Credit line, e.g. `Digitised by Google Books.`",
-          "type": "string"
-        },
-        "note": {
-          "type": "string"
-        },
-        "scan": {
-          "description": "Rights status of the digitisation itself, distinct from the work.",
-          "type": "string"
-        },
-        "work": {
-          "description": "Rights status of the intellectual work, e.g. `PD-old-100-expired`.",
-          "type": "string"
-        }
-      },
-      "type": "object"
-    },
-    "Roles": {
-      "anyOf": [
-        {
-          "type": "string"
-        },
-        {
-          "items": {
-            "type": "string"
-          },
-          "type": "array"
-        }
-      ],
-      "description": "`role = \"engraver\"` or `role = [\"surveyor\", \"draughtsman\"]`."
-    },
-    "Scan": {
-      "additionalProperties": false,
-      "description": "The digitisation event and its container file.\n\nThis is the one table with **selective** key inheritance: `file`, `by` and `url` inherit;\n`count` and `note` do not. See [`crate::load`].",
-      "properties": {
-        "by": {
-          "description": "Digitising agent, e.g. `Google Books`. Inherits.",
-          "type": "string"
-        },
-        "count": {
-          "description": "Number of pages in the container. Used by check 5. Does not inherit.",
-          "format": "int64",
-          "type": "integer"
-        },
-        "file": {
-          "description": "Path to the container, relative to the directory of the file that **declared** it.\nInherits.",
-          "type": "string"
-        },
-        "note": {
-          "description": "Caveats. Does not inherit.",
-          "type": "string"
-        },
-        "ppi": {
-          "description": "Resolution the scan was made at, in pixels per inch. Inherits.\n\nStated rather than inferred: `render` otherwise has to guess a page's native\nresolution from the largest image embedded in it.",
-          "format": "int64",
-          "type": "integer"
-        },
-        "url": {
-          "description": "Landing page for the digitisation, distinct from top-level `url`. Inherits.",
-          "type": "string"
-        }
-      },
-      "type": "object"
-    },
-    "Text": {
-      "additionalProperties": false,
-      "description": "OCR or transcription. Never inline content.",
-      "properties": {
-        "by": {
-          "description": "Producing agent and version, e.g. `tesseract 5.3 fra`.",
-          "type": "string"
-        },
-        "file": {
-          "description": "Path to the text file, resolved against the declaring file's directory.",
-          "type": "string"
-        },
-        "kind": {
-          "description": "`ocr` or `transcription`. Defaults to `ocr` when omitted.",
-          "type": "string"
-        },
-        "lang": {
-          "type": "string"
-        },
-        "note": {
-          "type": "string"
-        }
-      },
-      "required": [
-        "file"
-      ],
-      "type": "object"
-    },
-    "Zone": {
-      "additionalProperties": false,
-      "description": "A rectangle on a page, in the pixel space of that page's **primary graphic**.\n\nZones are not addressable and not citable; nothing in the address grammar reaches them.",
-      "properties": {
-        "id": {
-          "description": "Page-local label. Not a global id, not addressable.",
-          "type": "string"
-        },
-        "label": {
-          "type": "string"
-        },
-        "lrx": {
-          "description": "Lower-right x, exclusive.",
-          "format": "int64",
-          "type": "integer"
-        },
-        "lry": {
-          "description": "Lower-right y, exclusive.",
-          "format": "int64",
-          "type": "integer"
-        },
-        "note": {
-          "type": "string"
-        },
-        "ulx": {
-          "description": "Upper-left x, pixels, origin top-left.",
-          "format": "int64",
-          "type": "integer"
-        },
-        "uly": {
-          "description": "Upper-left y.",
-          "format": "int64",
-          "type": "integer"
-        }
-      },
-      "required": [
-        "ulx",
-        "uly",
-        "lrx",
-        "lry"
-      ],
-      "type": "object"
-    }
-  },
-  "$schema": "https://json-schema.org/draft/2020-12/schema",
-  "description": "One TOML file in the primary source and scan archive. Generated from the Rust types in tools/src/model.rs by `scans schema` - do not edit by hand.",
-  "oneOf": [
-    {
-      "additionalProperties": false,
-      "description": "The publication or work, independent of any copy.",
-      "properties": {
-        "country": {
-          "description": "Does not inherit.",
-          "type": "string"
-        },
-        "covers": {
-          "description": "EDTF interval. The span this source covers. Does not inherit.\n\nThis is where the legacy `held = \"1789\"` field lands.",
-          "type": "string"
-        },
-        "date": {
-          "description": "EDTF. Date of the work, for single-work sources.",
-          "type": "string"
-        },
-        "dimensions": {
-          "$ref": "#/$defs/Dimensions",
-          "description": "Physical size of the object. Does not inherit."
-        },
-        "extent": {
-          "description": "Bibliographic collation, `\"7 p. ; 18 cm.\"`. Does not inherit.",
-          "type": "string"
-        },
-        "founded": {
-          "description": "EDTF. Date the publication began.",
-          "type": "string"
-        },
-        "frequency": {
-          "description": "`daily`, `weekly`, … Free string.",
-          "type": "string"
-        },
-        "holding": {
-          "$ref": "#/$defs/Holding"
-        },
-        "id": {
-          "description": "Stable flat id. Defaults to the filename stem.",
-          "type": "string"
-        },
-        "identifier": {
-          "additionalProperties": {
-            "type": "string"
-          },
-          "type": "object"
-        },
-        "language": {
-          "description": "BCP-47 / ISO-639-1 code. Inherits.",
-          "type": "string"
-        },
-        "layer": {
-          "const": "source",
-          "type": "string"
-        },
-        "link": {
-          "items": {
-            "$ref": "#/$defs/Link"
-          },
-          "type": "array"
-        },
-        "material": {
-          "description": "What it is made of: `paper`, `laid paper`, `parchment`, `vellum`. Inherits.\n\nA list because a single object can be more than one — a paper text block in a vellum\nbinding. Feeds a Wikidata `made from material` claim (P186).",
-          "items": {
-            "type": "string"
-          },
-          "type": "array"
-        },
-        "note": {
-          "type": "string"
-        },
-        "of": {
-          "description": "Id of an ancestor. A source may be `of` another source (e.g. a series).",
-          "type": "string"
-        },
-        "page": {
-          "description": "Legal on a source when copy and document are collapsed in (Turgot: 21 pages).",
-          "items": {
-            "$ref": "#/$defs/Page"
-          },
-          "type": "array"
-        },
-        "pages": {
-          "$ref": "#/$defs/PageRange",
-          "description": "Terse page range. Legal but unusual on a source."
-        },
-        "place": {
-          "description": "Place of publication. Inherits.",
-          "type": "string"
-        },
-        "resp": {
-          "description": "Inherits wholesale. `Option` rather than a bare `Vec` because `resp = []` is the\ndocumented way to *clear* an inherited value, and that must be distinguishable from\n\"not declared here\".",
-          "items": {
-            "$ref": "#/$defs/Resp"
-          },
-          "type": "array"
-        },
-        "rights": {
-          "$ref": "#/$defs/Rights"
-        },
-        "scan": {
-          "$ref": "#/$defs/Scan",
-          "description": "Legal on a source when the copy layer is collapsed in (Turgot, Verniquet)."
-        },
-        "short_title": {
-          "type": "string"
-        },
-        "subject": {
-          "description": "Subject headings as the cataloguer wrote them. Does not inherit.",
-          "items": {
-            "type": "string"
-          },
-          "type": "array"
-        },
-        "text": {
-          "items": {
-            "$ref": "#/$defs/Text"
-          },
-          "type": "array"
-        },
-        "title": {
-          "type": "string"
-        },
-        "type": {
-          "description": "Genre: `newspaper`, `map`, `diary`, … Open vocabulary, not validated.",
-          "type": "string"
-        },
-        "url": {
-          "description": "Canonical landing page. Does not inherit.",
-          "type": "string"
-        }
-      },
-      "required": [
-        "layer",
-        "title"
-      ],
-      "type": "object"
-    },
-    {
-      "additionalProperties": false,
-      "description": "One physical object that was digitised.\n\nThe inner type is named `CopyRecord` rather than `Copy` so it cannot be confused with\nthe `std::marker::Copy` trait in type position. The serde tag is taken from the\n*variant* name, so the TOML still reads `layer = \"copy\"`.",
-      "properties": {
-        "country": {
-          "type": "string"
-        },
-        "covers": {
-          "description": "EDTF interval. The date span of material bound in this object. Used by check 6.",
-          "type": "string"
-        },
-        "date": {
-          "description": "EDTF. Date of the object itself, if meaningful.",
-          "type": "string"
-        },
-        "dimensions": {
-          "$ref": "#/$defs/Dimensions",
-          "description": "Physical size of the object. Does not inherit."
-        },
-        "extent": {
-          "description": "Bibliographic collation, `\"7 p. ; 18 cm.\"`. Does not inherit.",
-          "type": "string"
-        },
-        "holding": {
-          "$ref": "#/$defs/Holding",
-          "description": "The normal home for `holding`."
-        },
-        "id": {
-          "type": "string"
-        },
-        "identifier": {
-          "additionalProperties": {
-            "type": "string"
-          },
-          "type": "object"
-        },
-        "language": {
-          "type": "string"
-        },
-        "layer": {
-          "const": "copy",
-          "type": "string"
-        },
-        "link": {
-          "items": {
-            "$ref": "#/$defs/Link"
-          },
-          "type": "array"
-        },
-        "material": {
-          "description": "What it is made of: `paper`, `laid paper`, `parchment`, `vellum`. Inherits.\n\nA list because a single object can be more than one — a paper text block in a vellum\nbinding. Feeds a Wikidata `made from material` claim (P186).",
-          "items": {
-            "type": "string"
-          },
-          "type": "array"
-        },
-        "note": {
-          "type": "string"
-        },
-        "of": {
-          "description": "Id of the source (or another copy). Absent = orphan copy; legal, warns `W107`.",
-          "type": "string"
-        },
-        "page": {
-          "items": {
-            "$ref": "#/$defs/Page"
-          },
-          "type": "array"
-        },
-        "place": {
-          "type": "string"
-        },
-        "resp": {
-          "items": {
-            "$ref": "#/$defs/Resp"
-          },
-          "type": "array"
-        },
-        "rights": {
-          "$ref": "#/$defs/Rights"
-        },
-        "scan": {
-          "$ref": "#/$defs/Scan",
-          "description": "The normal home for `scan`."
-        },
-        "short_title": {
-          "type": "string"
-        },
-        "subject": {
-          "description": "Subject headings as the cataloguer wrote them. Does not inherit.",
-          "items": {
-            "type": "string"
-          },
-          "type": "array"
-        },
-        "text": {
-          "items": {
-            "$ref": "#/$defs/Text"
-          },
-          "type": "array"
-        },
-        "title": {
-          "type": "string"
-        },
-        "type": {
-          "description": "Physical genre: `volume`, `roll`, `folder`. Open vocabulary.",
-          "type": "string"
-        },
-        "url": {
-          "type": "string"
-        }
-      },
-      "required": [
-        "layer",
-        "title"
-      ],
-      "type": "object"
-    },
-    {
-      "additionalProperties": false,
-      "description": "The citable intellectual unit.",
-      "properties": {
-        "country": {
-          "type": "string"
-        },
-        "covers": {
-          "description": "EDTF interval, for documents spanning a range.",
-          "type": "string"
-        },
-        "date": {
-          "description": "EDTF. Used by check 6.",
-          "type": "string"
-        },
-        "dimensions": {
-          "$ref": "#/$defs/Dimensions",
-          "description": "Physical size of the object. Does not inherit."
-        },
-        "extent": {
-          "description": "Bibliographic collation, `\"7 p. ; 18 cm.\"`. Does not inherit.",
-          "type": "string"
-        },
-        "holding": {
-          "$ref": "#/$defs/Holding"
-        },
-        "id": {
-          "type": "string"
-        },
-        "identifier": {
-          "additionalProperties": {
-            "type": "string"
-          },
-          "type": "object"
-        },
-        "language": {
-          "description": "Inheritable scalar; present here so a document can override an ancestor's value.",
-          "type": "string"
-        },
-        "layer": {
-          "const": "document",
-          "type": "string"
-        },
-        "link": {
-          "items": {
-            "$ref": "#/$defs/Link"
-          },
-          "type": "array"
-        },
-        "material": {
-          "description": "What it is made of: `paper`, `laid paper`, `parchment`, `vellum`. Inherits.\n\nA list because a single object can be more than one — a paper text block in a vellum\nbinding. Feeds a Wikidata `made from material` claim (P186).",
-          "items": {
-            "type": "string"
-          },
-          "type": "array"
-        },
-        "no": {
-          "description": "Issue/serial number. Used by check 10.",
-          "format": "int64",
-          "type": "integer"
-        },
-        "note": {
-          "type": "string"
-        },
-        "of": {
-          "description": "Id of **any** ancestor — copy, source, or another document.\nAbsent = standalone document (the one-off engraving case).",
-          "type": "string"
-        },
-        "page": {
-          "items": {
-            "$ref": "#/$defs/Page"
-          },
-          "type": "array"
-        },
-        "pages": {
-          "$ref": "#/$defs/PageRange",
-          "description": "Terse page range."
-        },
-        "place": {
-          "description": "Inheritable scalar; present here so a document can override an ancestor's value.",
-          "type": "string"
-        },
-        "resp": {
-          "items": {
-            "$ref": "#/$defs/Resp"
-          },
-          "type": "array"
-        },
-        "rights": {
-          "$ref": "#/$defs/Rights"
-        },
-        "scan": {
-          "$ref": "#/$defs/Scan"
-        },
-        "short_title": {
-          "type": "string"
-        },
-        "subject": {
-          "description": "Subject headings as the cataloguer wrote them. Does not inherit.",
-          "items": {
-            "type": "string"
-          },
-          "type": "array"
-        },
-        "supplement_to": {
-          "description": "Id of another **document**. Check 8.",
-          "type": "string"
-        },
-        "text": {
-          "items": {
-            "$ref": "#/$defs/Text"
-          },
-          "type": "array"
-        },
-        "title": {
-          "type": "string"
-        },
-        "type": {
-          "description": "`issue`, `supplement`, `sheet`, `letter`, `play`, `engraving`, … Open vocabulary.",
-          "type": "string"
-        },
-        "url": {
-          "type": "string"
-        }
-      },
-      "required": [
-        "layer"
-      ],
-      "type": "object"
-    }
-  ],
-  "title": "Scan archive record"
-}

sources/actes-des-apotres/1789/11/actes-des-apotres-no-001.p1.txt

@@ -1,18 +0,0 @@
-LES ACTES
-
-DES A P OTR E S.
-
-Prîvatus illis fenfus erat brevis
-Commune magnum.
-
-<D
-
-'f*
-
-Lffifa'tf, J/éf- J^fi^U^-
-
-c
-
-#$.
-
-TQTHSCA

sources/actes-des-apotres/1789/11/actes-des-apotres-no-001.p2.txt

@@ -1,3 +0,0 @@
-^.s \
-
-4)C

sources/actes-des-apotres/1789/11/actes-des-apotres-no-001.p3.txt

@@ -1,30 +0,0 @@
-LES ACTES
-
-DES APOTRE &
-
-Chapitre Premier,
-
-.LiE bruit s'étant répandu que M. de Barentîn ;
-ancien garde- des -fceaux, était caché dans le
-couvent des Annonciades ., déguifé en novice y
-M. le chevalier de Lameth , membre du comité
-des recherches de l'AfTemblée Nationale j s'efl:
-tranfporté aux Annonciades , accompagné de
-150 hommes de la garde nationale parifienne.
-Le détachement s'efl: emparé de toutes les iflues;
-on a fait une vifite exacte du couvent , dans
-lequel M. de Barentin ne s'efl: point trouvé.
-
-On ne lauroic donuci. trpp d'éloges aux dlfpo-
-étions de M. de Lameth dans cette expédition :
-toutes les précautions avoient été prifes pour en
-aïîurer le fuccès , la vifite du couvent s'efl faite
-avec la décence que le lieu exigeoit , cV M.
-de Lameth s'efl; retké en ordre fans avoir perdu
-un feul homme.
-
-Des hommes connus en France fous le nom
-de perfifleurs, ont eflayé de répandre du ridi-
-cule fur la démarche de M. de Lamerh. Nous
-croyons rendre un fervice important à la patrie^
-en lui dénonçant le perfiflage comme une arifta-

sources/actes-des-apotres/1789/11/actes-des-apotres-no-001.p4.txt

@@ -1,31 +0,0 @@
-(4)
-
-cratîe, & de I'efpece la plus dangereufe ; car'o»
-peut définir le perfiflagc 3 l'ariftocratie de l'efprit.
-
-Tous les bons citoyens n'ont vu dans la con-
-duite de M. de Latneth que celle d'un patriote
-pur,, zélé, qui veille fans relâche au falut de la,
-chofe publique. Cette conduite a été fourenue
-depuis le commencement de l'AfTembiée. MM.
-de Lameth , iiîus d'une famille ariftocratique ,
-comblés des faveurs de la cour & des bontés
-de la Reine j n'ont ceiîé de fe montrer les
-plus arders défenfeurs de la démocratie, c'eft-
-â - dire , de la feule forme de gouvernement
-qui convienne à un grand empire. MM. de
-Lameth ontfacrifié lareronnaiifance particulière
-a Tintérêt public, & nous avançons fans crainte
-d'être contredits par les amis de notre liberté,
-que la non-gratitude, qui eft prefquc toujours
-le défaut des âmes viles , eft la vertu de MM.
-de Lameth.
-
-Nous ne connaifîons qu'un homme dans l'Af-
-femblée., qui ait autant de droits que MM. de
-Lameth à l'eftime & au refpcd de la Nation :
-Ôc dut notre franchife nous rendre odieux à tous
-ceux qu'on appelle alTez incorrectement les hon-
-nêtes gens , nous cédons au fentimer.t qui nous
-preffe > en rendant un hommage public aux ver-
-tus & aux talens de M. le comte de Mirabeau.

sources/actes-des-apotres/1789/11/actes-des-apotres-no-001.p5.txt

@@ -1,32 +0,0 @@
-(5 )
-
-Nous n'avons pu voir fans indignation l'auteur
-
-d'un libelle atroce refufer à M. de Mirabeau
-les talens même de l'écrivain. Pour décruire une
-aiïertionaufli injufte., nous oppoferons l'anonyme
-à lui-même. Il ne concerte point à M. De Mira-
-beau le Projet de Loi Martiale > ôc il vou droit
-nous faire croire que YAdrejJe aux Commettant
-eft de M. du Roveray. Nous conviendrons avec
-lafranchife dont nous faifons profeflion, qu'on
-trouve quelques incorrections de ftyie dans le
-Projet de Loi Martiale de M. de Mirabeau ;
-mais il eft convenu auffi en littérature d'appeler
-les incorrections les écarts du génie. Puifqu'il y
-a du génie dans l'Adreiîe aux Commettans , &c des
-incorrections dans le Projet de Loi Martiale, il
-en réfulte évidemment que l'un & l'autre font
-de M. de Mirabeau. Tous les bons citoyens
-feront de notre avis fur les ouvrages de ce grand
-homme , comme tous les bons logiciens le fe-
-ront fur (es vertus. Celle qui le diftingue peut-
-être de tous les hommes à talens qui brillent
-,dans FAflemblée ., eft fa modeftie ; c'eft le foin
-qu'il prend de dérober au public des adions hé-
-roïques. Nous devons le faire jouir déboute fa
-gloire , & publier , malgré lui , ce qu'il veut qu'on
-
-ignore.
-
-Le Vendredi 30 octobre, M. le comte de

sources/actes-des-apotres/1789/11/actes-des-apotres-no-001.p6.txt

@@ -1,28 +0,0 @@
-Mirabeau eut en fortant de l'AlTemblée une dif-
-euffion afïèz vive avec M. Cochercl , député de
-Saint-Domingue, fur Timpacience cV riiumeur
-qu'une partie de I'AiTemblée avait témoignées
-pendant les opinions de quelques honorables
-membres. Apres avoir échangé quelques far-
-cafmes affez durs avec fon adverfaire, le rarou-
-che américain, fe livrant à l'impétuoiué de fon
-caractère , s'emporta jufqu'a la menace, & pro-
-pofa à fon illuftre confrère de fortir. Plusieurs
-prélats , entr'autres M. l'évêque de Chalons-fur-
-Marne, conjuraient M. de Mirabeau de reftec
-« Eh , Mefîleurs ! s'écria M. Cocherel , épargnez-
-» vous tant de foins cV de peine, je vous réponds
-» qu'il ne fonge pas à fortir ».
-
-Croirair-on que nous avons trouvé beaucoup
-de gens dans le monde qui ont ofé taxer de
-lâcheté la modération & la prudence de M. de
-Mirabeau? Sa conduite nous paraît , au contraire,
-fort fage & très-méthodique ; elle s'explique
-d'elle-même. M. de Mirabeau a très-bien jugé
-qu'un combat entre deux députés ne ferait que
-fcandaleux & deviendrait interminable ; car deux
-perfonngs facrees & inviolables font nécessaire-
-ment invulnérables. Voilà le feul motif qui a
-enchaîné fon courage. Ils n'en douteront pas ceux
-qui om été à portée comme nous de vo.ir cet ex-»

sources/actes-des-apotres/1789/11/actes-des-apotres-no-001.p7.txt

@@ -1,33 +0,0 @@
-(7 )
-
-cellent citoyen fe promenant le lundi 5 octobre
-fur la place d'armes à Versailles , fon fabre nud
-fous le bras , avec le maintien qui caradtérife le
-vrai courage , éc qui lui valut cet éloge flatteur
-de la part du commandant du régiment de Flan-
-dres : « Vous nous retracez Charles XII .
-
-Ce mot laconique dans la bouche d'un brave
-& ancien officier n'eft point le langage de la flat-
-terie j c'eft Texpreflion pure de la vérité & du
-fentiment d'admiration qu'infpirent les plus pe-
-tites actions de M. de Mirabeau , & même fon
-inaction ( 1 ).
-
-Français , peuple fenfible & généreux 9 fi digne
-de la liberté ! ne méconnaifTez jamais vos bien-
-faiteurs ! que vos cœurs ne s'occupent que de la
-
-( 1 ) Croirait - on qu'il y a encore des ariftocrates
-qui fe font fâchés qu'un vieux paladin , le comte de
-Lautrec , rencontrant fon confrère Mirabeau le lende-
-main dans le veftibule de TAfTemblée Nationale, lui
-fît fon compliment, croyant fans doute que le Créole
-était tué ; mais fur l'obfervation qu'on lui fit que le
-tribunal du point d'honneur étant fupprimé, le point
-d'honneur ne pouvait plus exifter , le vieux chevalier
-convint de fon tort , mais provifoirement , il retira
-fon compliment. Nous attendons la dénonciation que
-JVL de Mirabeau fera probablement de ce fait, après
-les 157 dénonciations quil a annoncées avant de pro->
-poncer notre jugement définitif.

sources/actes-des-apotres/1789/11/actes-des-apotres-no-001.p8.txt

@@ -1,29 +0,0 @@
-(*)
-
-reconnailTance que vous leut devez , ôc qu'ils
-(oient toujours fermés aux craintes , aux foup-
-çons que la calomnie voudrait y répandre! que
-cette reconnailTance ne foit point ftérile , ôc ne
-laiffez point à la poftérité le foin tardif d'acquitter
-envers vos immortels législateurs la dette de la
-Patrie! Songez que les grands hommes appar-
-tiennent à l'univers, & que la France va fe cou-
-vrir d'une honte éternelle, fi elle ne s'empreffe
-de confacrer les noms de ceux qui ont fondé fa
-liberté. La Grèce eut fes Lycurgues, fes Solons;
-Rome fon Numa ; la France aura fes Cottin Ôc
-fes Mirabeau, fes Robertspierre & Duport, fes
-Lameth ôc (es Chapelier , Target & fes Corol-
-lers y Brouftaret & Barnave , fes Lapoule & fes
-Labefte , fes Champeaux , fes Populus , fes Ker-
-velegans ôc tant d'autres qui font l'étonnemenc
-de l'Europe attentive.
-
-Nous n'ouvrons pas une fouÇcription pour nos
-Actes des Apôtres. Il ejî pojjîble que la matière
-foit ajfei abondante pour livrer tous les jours un
-chapitre à Pimprejjion ; mais nous ne voulons pas
-prendre d'engagement : nous connoijjons notre
-pareffe , £r nous ne /avons pas jujqu'oà peut
-aller VaBïvitè héroïque des honorables membres
-de rJJJemblée,

sources/actes-des-apotres/1789/11/actes-des-apotres-no-001.toml

@@ -1,73 +0,0 @@
-#:schema ../../../../schemas/source.json
-id    = "actes-des-apotres-no-001"
-layer = "document"
-of    = "actes-des-apotres-vol1"
-type  = "issue"
-no    = 1
-date  = "1789-11?"
-pages = { from = 7, to = 14 }
-note  = "Start fixed by its own heading, `Chapitre` and the number spelled out; the range ends at the title leaf of the number after it. 8 pages: title leaf, blank verso, then the text from printed page 1. Dates named inside it, day/month: 5/10, 30/10. The month is inferred, not printed - see the source record."
-
-[[page]]
-n = 1
-[[page.text]]
-file = "actes-des-apotres-no-001.p1.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 2
-[[page.text]]
-file = "actes-des-apotres-no-001.p2.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 3
-[[page.text]]
-file = "actes-des-apotres-no-001.p3.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 4
-[[page.text]]
-file = "actes-des-apotres-no-001.p4.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 5
-[[page.text]]
-file = "actes-des-apotres-no-001.p5.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 6
-[[page.text]]
-file = "actes-des-apotres-no-001.p6.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 7
-[[page.text]]
-file = "actes-des-apotres-no-001.p7.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 8
-[[page.text]]
-file = "actes-des-apotres-no-001.p8.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p1.txt

@@ -1,5 +0,0 @@
-LES ACTES
-
-DES APÔTRES
-
-Efurientcs implevit bonis & dirttes d'-mijït inanes.

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p10.txt

@@ -1,35 +0,0 @@
-( I»)
-
-Elizabeth eut pris le fécond nom du royaume ,
-le nom de Madame le Chapelier dans toutes
-les formes.
-
-Que cela ferve à confondre ceux oui, trom-
-pés par les lunettes qu'il porte habituellement ,
-croient que M. Chapelier n'a pas des vues
-étendues.
-
-M. Barnave avait, dit-on, demandé la charge
-de prévôt -général des maréchaulTées du royau-
-me , & fa tendre jeunelTe nous promettait de
-longs & utiles fuccès.
-
-M. Populus s'était , à ce qu'on aflure a ré-
-fervé les haras»
-
-M. Cottin , l'intendance de nos colonies ôc
-la législation future des gens de couleur ,
-auxquels il prend, dit-on, un très-vif intérêt»
-
-M. Lavenue était déjà nommé grand-maître
-des eaux ., bois & forêts.
-
-Un fceau national nous refte encore à créer;
-un membre tout frais fe préfentait , & c'était
-Duport , qui alloit tenir en échec tous nos
-tigres & tous nos ogres de la chicane.
-
-M. le Comte de Mirabeau n'avait aucun dé*
-partement ; également propre â tous j fers la
-guerre, il eût plané fur tous., fans en diriger
-aucua en particulier ; tel que cette Divine Pro-

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p11.txt

@@ -1,30 +0,0 @@
-( » )
-
-vidence qui veille fur l'efpece humaine fans
-être apperçue de perfonne.
-
-Une fuite de ces maux qui nous accablent;
-a encore fait avorter des projets fi heureufemenc
-conçus. Qu'arrivera-til de-là ? que nous man-a
-querons de tout cet hiver , & fans dourede pain;
-
-M. le Comte de Mirabeau avait cependant
-pourvu à tous nos befoins depuis long - tems.
-Dès le 5 feptembre , un de (es amis intimes,
-le fieur Fauconnier ^ un des plus grands négo-
-cians de l'Europe , aufîi grand capitalifte que
-profond fpéculateur , s'était rendu à Londres,
-& avait fait l'emplette de 16000 facs de fa-
-rine ôc 11000 facs de bleds pour Paris. Les mem-
-bres envoyés dans le courant du même moif
-par la municipalité , n'y trouvèrent plus de
-fubfiftances à acheter ; mais ils rapportèrent
-Tefpoir de précéder de peu de jours les achats
-immenfes de l'ami de M. le Comte de Mira-
-beau. Un démon envieux n'a-t-il pas voulu que
-des banquiers de Paris très-puniflablesjettaiTent
-du trouble & répandifient des doutes fut le
-crédit qu'a toujours mérité M. Fauconnier. Au
-moyen de ce , les grains font toujours reftés à
-Londres. Ces banquiers feront certainement
-dénoncés quelque jour, & M. de Mirabeau,

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p12.txt

@@ -1,31 +0,0 @@
-" )
-
-le Drprcmefnil du parlement national ., doit
-cette étude à la nation.
-
-Un nouveau travail dont le comité des finances
-va avoir à s'occuper, fera relatif aux indem-
-nités dues à tous les fourni fleurs , écuyers , ta-
-piffiers , feliiers , maîtres d'hôtel , &C. qui pré-
-paraient d jà la ma ton miniftérieile de M. le
-Comte. On y aura certainement égard, un
-décret ne peut pas avoir d'effet rétroactif.
-
-Il en fera de même de rous les fecretaires,
-• Fers ci mroh J u tendans , cVc. qui vont fe
-trouver défapcknntés , c'eft une nouvelle charge
-pour la caiife nationale. Il faut favoir faire des
-facrihees.
-
-Ne défefperons cependant point du falut pu-
-blic tant que M. le Comte corjfervera fon in-
-fluence dans I aifemblce. Nous avons déjà eu
-en quelque force le renou ■• ellement de la jour-
-née des barricades , des m pîaifans nom-
-ment . - famedi 7 , la journée des dupes;
-ne doutons pas que ce ptince patriote qu'on a (1
-indignement déchiré dans un libelle atroce j
-n'arrive de Baril au premier jour avec M. Laclos
-& M. Duroyeray , fes deux aide rp , a
-la tête d'un c nvoi de fubiiftances , èx* nou>
-aurons prefque au nie

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p13.txt

@@ -1,27 +0,0 @@
-( M )
-
-jbarricides , la journée des dupes & la journée
-des farines. Quels matériaux pour l hiftbire !
-
-Les deux motions de M. le Comte de Mi-
-rabeau far les américains ., 'S: ia cailfe d'ef-
-compte avait et: ajournées j nous en parlerons
-une autre fois.
-
-Il vient de nous tomber fous la main l'extrait
-d'une lettre de M. le comte de Lally TolendaL
-oui ciraile dans Paris ; nous; nous emprelTons
-de la publier po.u: montrer j.ïfqu'où peut aller
-la fureur & le délire : encore un homme qui
-ofe inculper M. de Mirabeau , comme (i dans
-l'adrefife à M. Baiily &■ la Fayette, rédigée
-par lui lors de la tranîlation de l'alfemblée à
-Paris , il ne s'était pas fuffifamment juitihé aux
-yeux des bons citoyens, endifant que dans les
-grandes occalions il écoic nécetfaire de prendre
-jufqu'à l'attitude de con (pirateurs.
-
-11 eft à délirer que l'on découvre l'original
-de cette lettre , afin d en ponrfuivre l'écrivain
-quand on aura, fait des loix fur la liberté des
-. lettres , ôc fur celle de la preile.

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p14.txt

@@ -1,28 +0,0 @@
-( 14 ï
-
-g -, ai ■ >-*.
-
-EXTRAIT
-
-D'une lettre de M de Lallx - Tolendal*
-
-» Parlons du parti que j'ai pris , il eft bien
-y> juftifié dans ma confeience. Ni cette ville
-«• coupable , ni cette aflemblée plus coupable
-» encore ( i ) ne mériteroient pas' que je me
-» juftifié ; mais j'ai à cœur que vous & les
-» perfonnes qui penfent comme vous , ne me
-« condamnent pas. Vous devez déjà m 'avoir
-a lu , & par ce que j'ai dit , avoir jugé ce que
-» je taifois. Ma fanté , je vous jure, me ren-
-s> drait mes fondions imposables ; mais même
-m en la mettant de coté , il a été au-deflus de
-» mes forces de fupporter plus long-tems l'hor-
-» reur que me caufaient ce iang, ces têtes*
-55 cette reine prefqu'égorgée , ce roi amené
-» efclave , entrant à Paris au milieu de ces
-» aflaflins , & précédé des têtes de ces mal-
->» heureufes gardes. Ces perfides janniflaires >
-
-( i ) C'eft une nouvelle figure de rhétorique , par la-
-quelle on emploie le tout pour h partie.

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p15.txt

@@ -1,26 +0,0 @@
-» ces afTaflins ; ces femmes cannibales ; ce ctï
-» de tous les évêques à la lanterne , dans le
-a> moment où le Roi encre en fa capitale avec
-m deux évêques de fon confeii dans fa voiture ,
-a un coup de fufil que j'ai vu tirer dans un
-» dos carroffes de la Reine , M Baiily appel-
-3> lant cela un beau jour ; TafTemblée ayant dé-
-» claré froidement le matin, qu'il n'était pas
-« de fa dignité d'aller toute entière environnée
-v> le Roi ; M. de Mirabeau difant impuné-
-» ment dans cette afiemblée ., que le vaifleau
-« de l'état > loin d'être arrêté dans fa courfe ,
-» s'élançait avec, plus de rapidité que jamais
-a vers la régénération. M. Barnave riant avec
-» lui quand des flots de fang coulaient autour
-m de nous. Le vertueux Mounier échappant
-» par miracle à 20 aflTaiîins qui avaient voulu
-» faire de fa têce un trophée de plus. Voili
-» ce qui me fit jurer de ne plus mettre le
-» pied dans cette caverne d'antropophages, où
-aa je n'avois plus la force d'élever la voix , où
-b depuis fix femaines je Pavois élevée j envain ,
-y> moi , Mounier , & tous les honnêtes gens ,
-» ou le dernier effort à faire pour le bien était
-» d'en fortir • aucune idée de crainte ne s'eft
-» approchée de moi j je rougirais de jïïen dé-

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p16.txt

@@ -1,28 +0,0 @@
-( 16 )
-
-e. J'avais encore reçu for la rrure de
-« la part Je ce peuple moins coupable que
-m ceux qui. Tout enivré de fureur , des acla-
-» mations 8c des applaudi (îetnen s dont d'au-
->» î ,ent âtc & qui m'ont fait
-
-D3 frémir. C'eft à l'indignation , c'eft à une
-,j honeur irréfiftible , c'eft aux convulfions phy-
-t> fiques y que ce feul afpecfc me faifaic éprouver
-53 que j'ai cédé; on brave une feule mort, on
-3> la brave plufïeurs fois , quand elle peut être
-» utile; mais aucune puitTance fous le ciel >
-y> mais aucune opinion publique ou privée ,
-sa n'ont le droit de me condamner à fouffrir
-» inutilement mille fupplices par minute , Se
-s> à périr de défefpoir , de ra^e , ai milieu
-j> des triomphes du crime que je n'ai pu ar-
-d> rèter. Ils me preferiront, ils confifqueronc
-» mon bien , je labourerai la terre , & je ne
-*> les verrai plus. Voilà ma jnftification. Vous
-» pouvez la lire, la montrer, la laider copier,
-j> tant pis pour ceux qui ne la comprendront
-» par, ce ne fera a'ors que moi qui aurai eu torr
-u de la leur donner ».
-
-F I Ni

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p3.txt

@@ -1,27 +0,0 @@
-ttUL
-
-LES ACTES
-DES APÔTRES.
-
-Chapitre IL
-
-JSJoys comptions, en commençant nos Ac-
-tes, diriger tour- à- tour fur chacun de nos apô-
-tres l'artillerie de nos louanges. Une fatalité
-inconcevable nous contrarie à chaque moment^
-ôc nous détourne de notre marche. Un feui
-homme abforbe toutes nos facultés & nos refpedts.
-L'univerfalité des vertus ôc l'omni-fcience de
-M. le comte de Mirabeau nous forcent, malgré
-nous , à rappeler fans celTe l'attention de no»
-Lecteurs fur cet être admirable. Il femble réa-
-lifer les merveilles de la fable, Calchas ne per-
-çait pas mieux dans l'avenir j Prothée ne fe
-déguifait pas fous une plus grande quantité de
-formes, & de formes plus aimables j Alcibiade
-«tait moins modefte y & Fabricius n'exerça ja-
-mais à un pareil degré le défintérefiemenr, &
-le deftr d'être utile fans être remarqué. Si la,
-plupart de fes collègues font étrangers à tout »
-
-Ai

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p4.txt

@@ -1,30 +0,0 @@
-(4)
-
-il n'eft étranger à rien. Arithmétique politique;
-marine, finances, diplomatique-, agiiculture,
-population , en un mot , tout ce que le globe
-:re eft englobé dans fa tue vafte & pro-
-fonde ; une éloquence douce & fLurie parfume
-fes lèvres de roie , la perfuafn n marche à fa
-fuite ; une probité rare , une candeur aimab'e
-animent, colorent cet heureux enf mb'e; c'eft
-l'image du jwfie dont Horace nous tait le tableau;
-integer vit a fcelerumque purus. C'eft l'emblème
-de l'orateur accompli que Quintilien nous défi-
-nit : f^ir bonus d-xendi pultus.
-
-Pourquoi faut il que l'envie s'attache rcujours
-fur 1-S#pas des grands hommes? des farcîmes
-injurieux , de plats calembourgs , des libelles
-atroces , voilà les armes avec lefquelles des
-pygmées attaquent des géansjmais quand on a
-été trempé, corr.me M. le Comte., dans les eaux
-duSiyx_, tous ces tiaits s'émoufle».t devant fon
-invulnérabilité, & le grand homme, le héros
-intangible refte toujours i honneur de la géné-
-ration contemporaine & le moiele de généra-
-tions  futures.
-
-Nous devons des éxeufesà M-Jlieurs les mem-
-bres du pouvoir législatif cv exécutif, li le tribut
-des éioges qu'ils méritent ell li iouvenc arrêté j

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p5.txt

@@ -1,32 +0,0 @@
-( 5 >
-qu'ils fe montrent par des côtés aufia brillans
-que leur digue & rcfpeélabie modèle , ils par-
-tageront fa gloire j ils partageront no:re gra-
-tirude ; mais en attendant , nous les prions d'i-
-miter feu Lazare de fainte mémoire : qu'ils fe
-contentent des miettes qui tomberont de la table
-de notre mauvais riche.
-
-M. le Comte vient de faire trois motions .,
-toutes également recommandables , toutes éga-
-lement patriotiques , elles refpirent le civifme
-le plus fubtiî ; mais Tefprit de ténèbres a frappé
-d'aveuglement Meilleurs les députés. Enfin, on
-le croirait à peine, ce font fes Bretons, fe^ fidè-
-les Bretons qui l'ont attaqué le plus vigoureu-
-fement , & c'eft M. Blin, médecin de Nantes,
-qui lui a porcé le coup mortel. Il a pu dire ,
-en s'cnveloppant la tête ti mme le vainqueur
-du monde : Tu quoque mi Brute,
-
-On ne fait ce qu'on doit le plus admirer
-dans M. le Comte , de foq inconcevable facilité >
-de fa pénétration , ou de fa bienveillance pu-
-blique j les que (lions les plus ardues, les dif-
-euflions les plus longues &z les plus f.vantes ,
-tout cela ne lui coûte à concevoir que le tems
-de les lire.
-
-Les trois motions importantes qu'un jour a
-
-a3

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p6.txt

@@ -1,34 +0,0 @@
-m
-
-vu nîître Ôc mourir , avaient pour objet:
-
-i°. De nous faire payer des Américains en
-fubfiftances ce que nous leur avons avancés pour
-conquérir leur liberté & leur confîftanc*
-politique.
-
-20. De fupprimer la caiiïe d'efcompte., éta-
-blifTement ariftocratique, vieux , féodal , ufu*
-taire & banqueroutier.
-
-3°. D'obliger les miniftres à fe tenir à la barre
-de l'aiTemblée pour rendre à chaque inftant de
-la journée compte de ce qu'ils auroient , fans
-doute , fait la nuit.
-
-De mauvais citoyens fe font permis des far-
-cafmes & des obfer varions fi injurieufes à la
-réputation de M. de Mirabeau , que nous de-
-vons 5 pour fa gloire & notre repos , dévoiler
-les motifs fecrets qui ont dû l'animer en fai-
-fant ces trois proportions.
-
-Nous commençons par celle qui exigeait la
-préfence des miniftres.
-
-Il eft certain, & tous les gens profondément
-inftruits en conviennent, que M. Necker qui
-dirige à-peu-près le Confeil du Roi, a la tête
-ufée , qu'abforbé de chagrins , de longues &
-antiques médications, fts organes font épuifées ;
-on convient, à la vérité., qu'il a une grande

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p7.txt

@@ -1,33 +0,0 @@
-(?>■
-
-expérience du palTé; la fcience du prêtent 3 l*ar£
-de temporifer , un certain capital de vertu , da
-probité , même une forte de confiance au-dehots
-& au-dedans de la France ; mais que font toutes
-ces qualités devant la fublime fcience des futurs
-contingens , de cet art régénérateur de nos
-magiciens modernes qui , épuifant le vieux fang ,
-nous promettent de faire couler dans nos veines
-rajeunies un nouveau principe vital. Il eft vrai
-que, moins heureux qu'Efcn , nous pouvons
-fuccomber dans l'opération j mais ne vaut- il
-pas mieux mourir avec grâce > que vivre de
-plâtrages.
-
-M. Necker & tous les autres miniftres n'euf-
-fent certainement pas voulu conferver le mi-
-niftere à des conditions fi patriotiques 8c f*
-honorabies ; de mauvais plaifans euffent tourné
-en ridicule leurs fonctions nouvelles , ieut dé-
-miflion était au bout de la motion , ôc le nou-
-veau miniftere s'élevoit naturellement fur les
-débris de l'ancien. Virgr jeunes membres , £f*
-lue de la nation ^ arrivaient tout neufs aux af-
-faires., & la profpérité publique naifïbit de leur
-félicité particulière.
-
-Quel malheur pour la France qu'un pareil
-
-changement n'ait pas eu lieu , ôc que fous le
-
-A4

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p8.txt

@@ -1,29 +0,0 @@
-( 8 )
-
-vain prétexte de ne point entretenir dans une
-aflemblée, qui ne doit avoir qu'un efprit y
-un fentiment d'oppofnion ambitieufe , on aie
-déclaré que nul membre du corps législatif ne
-pourra erre miniftre. N'eft-ce pas détruire le
-germe de cette précieufe émulation dont l'An-
-gleterre a jufqu'ici retiré des min'iftres fi par-
-faits & fi probes \ il eft vrai qu'elle produit une
-continuité d'orages , d'intrigues , même des
-feenes fanglantes : ici eft la bannière de Fox,
-là celle de Pitt ; mais auflï exifte-t-il un ciei
-aufïi pur par intervalles que celui des Antilles
-après les inondations, les débordemens , les
-ouragans 6c les trembîemens de terre qui les
-défolent périodiquement tous les ans ?
-
-Avec quel plaifir la nation n'cût-elle pas vu
-fièger fur la fellette miniftérielle comme chan-
-celier de l'échiquier, Mgr. Tévêque d^utun. En
-vain de mauvais plaifans , s'égayant fur l'acci-
-dent qui le prive de Pufage de (es jambes,
-difent qu'il ne marchera jamais Gro|t. L'heu-
-reufe influence de M. le Comte l'eût bientôt
-redretfe , & la droiture de celui-ci nous répon-
-dait de la rectitude des démarches de (on pro-
-tégé. Pourquoi faur-il que Monfeigneur _, après
-avoir décompofe fi habilement iesf chances de

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.p9.txt

@@ -1,30 +0,0 @@
-(?)
-
-la loterie royale ^ fe foie fourvoyé dans ce
-tirage national? c'eft un malheur qu'il faut
-favoir fupporter. Monfeigneur , après fa mo-
-tion , a plus de force que nous pour prendre
-fon parti , c'eft un bénéfice de plus à réfigner ,
-il faura s'exécuter , & la pratique confirmera fa
-théorie.
-
-M. de Caftellane devoit, dit -on, avoir le
-département de Paris. Tous les honnêtes gens,
-les créanciers malheureux n'auraient pu man-
-quer d'applaudir à cette difpofition. Diftributeur
-des arrêts de furféance , M. de Caftellane ne
-s'en fût permis qu'un feul par année j fes prin
-cipes à cet égard font connus.
-
-M. le Chapelier n'avait pas encore (on dé-
-partement fixé. On aurait pu créer pour lui une
-charge de bâtonnier de tous les prefidens du
-royaume. Un ex préfident de l'allemblée natio-
-nale eft fans doute fupérieur à toutes les cours
-fouveraines de l'empire. Quand on a été la
-féconde perfonne de l'état après le fouverain >
-on ne peut plus rétrograder j aufïï', ajoutaît-on,
-que provifoirement on allait lui donner Mde.
-Elizabeth en mariage. On l'eût récompenfé
-ainfi d'avoir li dextrement mis toutes les têtes
-dans un bonnet la nuit du 4 acût , & Madame

sources/actes-des-apotres/1789/11/actes-des-apotres-no-002.toml

@@ -1,132 +0,0 @@
-#:schema ../../../../schemas/source.json
-id    = "actes-des-apotres-no-002"
-layer = "document"
-of    = "actes-des-apotres-vol1"
-type  = "issue"
-no    = 2
-date  = "1789-11?"
-pages = { from = 15, to = 30 }
-note  = "Start fixed by its title leaf: the headings on either side fix which number it is; the range ends at the title leaf of the number after it. 16 pages: title leaf, blank verso, then the text from printed page 1. The month is inferred, not printed - see the source record."
-
-[[page]]
-n = 1
-[[page.text]]
-file = "actes-des-apotres-no-002.p1.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 2
-
-[[page]]
-n = 3
-[[page.text]]
-file = "actes-des-apotres-no-002.p3.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 4
-[[page.text]]
-file = "actes-des-apotres-no-002.p4.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 5
-[[page.text]]
-file = "actes-des-apotres-no-002.p5.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 6
-[[page.text]]
-file = "actes-des-apotres-no-002.p6.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 7
-[[page.text]]
-file = "actes-des-apotres-no-002.p7.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 8
-[[page.text]]
-file = "actes-des-apotres-no-002.p8.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 9
-[[page.text]]
-file = "actes-des-apotres-no-002.p9.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 10
-[[page.text]]
-file = "actes-des-apotres-no-002.p10.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 11
-[[page.text]]
-file = "actes-des-apotres-no-002.p11.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 12
-[[page.text]]
-file = "actes-des-apotres-no-002.p12.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 13
-[[page.text]]
-file = "actes-des-apotres-no-002.p13.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 14
-[[page.text]]
-file = "actes-des-apotres-no-002.p14.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 15
-[[page.text]]
-file = "actes-des-apotres-no-002.p15.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"
-
-[[page]]
-n = 16
-[[page.text]]
-file = "actes-des-apotres-no-002.p16.txt"
-kind = "ocr"
-by   = "ABBYY FineReader 8.0"
-lang = "fr"

sources/actes-des-apotres/1789/11/actes-des-apotres-no-003.p1.txt

@@ -1,7 +0,0 @@
-:
-
-LES ACTES
-DES APOTRES,
-
-En dépit des méchans
-Je vois qu'on en revient toujours aux bonnes gens.

sources/actes-des-apotres/1789/11/actes-des-apotres-no-003.p10.txt

@@ -1,26 +0,0 @@
-( lo)
-
-& des rivières qu'un royaume doit fe divifer
-Une batterie bien portée , fervie par une lé-
-gion de Therfites , eût fait plier la phalange
-macédoniene ; Achile lui-même eût reculé de-
-vant le piftolet de Mirabeau. Ne longeons donc
-qu'a former une nouvelle tactique de gouver-
-nement , brifons les fhtues de Louis XII &
-d'Amboife, d'Henry IV Ôc Sully, & rendant
-mille actions de grâce aux modernes Vaubans
-de la politique, au nouveau Newton de la
-finance , remettons aux mains des mathémati-
-ciens le foin de nous gouverner , & pour afïurer
-à nos neveux les probabilités d'un bonheur dans
-o utes les règles > confions à M. de Condorcet l'édu-
-cation & le préceptorat de Mgr. le Dauphin.
-Apres cette digreffion qui nous a entrainés
-hors de notre fujet , nous devons en revenir
-à M. Necker. Nous avons remarqué avec plaifir
-que M. de Mirabeau a été feul infenûble aux
-applaudilTemens qui ont précédé & fuivi l'en-
-trée de ce premier dans la falle; Ci jufquici ,
-on a pu dire de M. de Mirabeau > illi robur &
-£s triplex, en contemplant famedi dernier fon
-attitude fevere, on pouvait ajouter peclus ahe-

sources/actes-des-apotres/1789/11/actes-des-apotres-no-003.p11.txt

@@ -1,28 +0,0 @@
-( 11 )
-
-nétum , auflî difait-il à fes collègues en' parlant
-de l'orateur genevois : Il y a long-tems que
-je le méprife autant que je le détefte ; nous
-le tenons , il ne nous échappera pas , je me
-charge de le pulvérifer. Ces heureufes difpo-
-fîtions ont ranimé notre efpoir ; M. Necker a
-effectivement l'habitude d'accaparer les hom-
-mages , on ne peut ni le voir ni l'entendre ,
-fans que l'univerfalité des bénédictions du peuple
-ne l'accompagne. On le voit, on efperej il
-parle > on applaudit ; il fe retire , on applaudie
-encore même fans l'avoir entendu. Un tel ci-
-toyen pourrait être fort dangereux , & M. de
-Mirabeau qui connaît à fond toutes les belles
-loix d'Athènes, peut faire encore un grand ufage
-de celle de l'Oftracifme.
-
-Les bons efprits n'ignorent pas qu'il exifte en
-Sui/Te des Louftiesqui ont le talent de dérider tout
-un régiment en faifant tire le chef de file. C'eft par
-un magnétifme à peu près femblable, que le public
-ayant vu M. àc Mirabeau rire fardoniquement
-après le difeours de M. Necker , s'eft mis d'un
-commun accord à fourire gracieufement au
-miniftre qui l'égaré & qui , il Ton n'y prend
-garde , va perdre la Fiance.

sources/actes-des-apotres/1789/11/actes-des-apotres-no-003.p12.txt

@@ -1,25 +0,0 @@
-Ce n'eft que vendredi prochain que cette
-intéreflante queftion fera débattue ; jufque la
-nous compterons les minutes : faifons des vœux
-pour que dans la nouvelle organifacion de-
-là banque raronale , nous n'ayons plus à redouter
-d'y voir toute certe armée de banquiers , finan-
-ciers j capitaliftes,efcompteurs , courtiers 5acl:io-
-niftes j agioteurs, négocians > mar hands , manu-
-facturiers dont nous avons tant à nous plaindre;
-mais puiflîons-nous y voir 24 philofophes , 24
-arithméticiens politiques , 24 publiciftes , voire
-même fans boucles & fans fouliers , que ce
-nouvel affemblage national préfidé par MM.
-de Mirabeau , Claviere, Duroveray , Briiïbt de
-Warville, nous confirment chaque jour cet adage
-de mon ami Figaro : il ncfl pas néceffaire de
-tenir tes chofes pour en raifonner , ou bien ce
-proverbe encore plus vrai de mon ancre ami
-Greffet :
-
-En dépit des méchans ,
-Je vois qu'on en revient toujours aux bonnes gens.
-
-Nous avons vu , après le départ de M.
-Necker, décider avec plaiiir la grande queftion

sources/actes-des-apotres/1789/11/actes-des-apotres-no-003.p13.txt

@@ -1,13 +0,0 @@
-des deux féances par jour crois fois la femaîne;
-cela prouve que nos repréfenrans dont l'eftomach
-fe fatiguait à attendre ju fqu'à 5 heures _, font
-difpofés à faire toutes fortes de facrifices pour
-notre bien , ils renoncent à l'opéra > pour mettre
-plus de fuite à leurs opérations : on a vu à
-Verfailles le 4 août , quelle étoit l'influence
-d'un bon dîner , & c'eft avec reconnai(Tance
-que nous apprenons a lapoftérîté que M. Pain,
-M. Perdrix , M Salle de Choux , & les deux
-MM. Fricot ont été de l'avis de la majorité.
-
-F I N.

sources/actes-des-apotres/1789/11/actes-des-apotres-no-003.p3.txt

@@ -1,27 +0,0 @@
-JX.*<fà>frjLLU , , , , y
-
-LES ACTES
-
-DES APOTRES.
-
-Chapitre Troisième.
-
-J^t o u s avons promis à nos Le&eurs de
-revenir fur la triple motion , faite le fix de
-ce mois par M, le Comte de Mirabeau ; nous
-avons déjà rendu compte des motifs qui l'avaient
-déterminé à demander la préfence continuelle
-des Miniftres à l'Afïemblée Nationale , pour y
-fubir interrogatoire toutes fois &quantesil aurait
-plu à M. Brouftaret, à M. de Peluzat, à M.
-Cottin , à M. Floc > à M. Buttafuoco , Sec. > de
-leur adreflTer la parole \ bous Confirmons nos re-
-grets de ce qu'une conception fi heureufe foit
-avortée , & qu'il nous faille renoncer pour trois
-ans à voir M. Duport en fimarre \ M. de Mi-
-rabeau avec fon bec à corbinj M. le Chappellier
-avec le bâton fous fon habit , pour la première
-fois y compofant, en face de la tribune, la tête
-de la ligne du Confeil d'Etat , & faifant l'exercice
-
-A i

sources/actes-des-apotres/1789/11/actes-des-apotres-no-003.p4.txt

@@ -1,30 +0,0 @@
-(4)
-
-au commandement de tous les anfpelfades na-
-tionaux. On a eu de bien faufles notions fur l'é-
-lévation du cœur humain ôc la dignité de l'hom-
-me, lorfqu'on a penfé qu'il ne fe trouverait au-
-cun membre, même dans l'Atfemblée, qui vou-
-lût Te ioumettre à une cérémonie qui retracerait
-en quelque forte l'ancienne queftion. Peu farni-
-liarifés encore avec les grandes idées du patrio-
-tifme , nous ne l'avons pas jufqu'où peut aller le
-dévouement à la chofe publique. Le facrifice que
-ces Meffieurs faifaienc fur l'autel de la patrie de
-leur dignité Ci connue ôc C\ admirée., fera la honte
-éternelle de la Nation qui l'a méconnue.
-
-La féconde motion de M. le Comte de Mira-
-beau avait pour objet d'établir une Caiife Na-
-tionale , qui reçut ôc payât la dette des créanciers
-de l'Etat, affermît le crédit, diminuât la dette
-publique j ôc remplaçât à jamais cet établifTe-
-ment impur de la rue Vivienne, nommé Caijfc
-d'Efcompte.
-
-Nous attendions avec une impatience fans
-bornes le jour heureux où l'AiTemblée dévoie
-prendre en confidération ce travail important,
-perfuadés que M. le Comte de Mirabeau allait
-
-Kz

sources/actes-des-apotres/1789/11/actes-des-apotres-no-003.p5.txt

@@ -1,31 +0,0 @@
-( 5 )
-
-nous présenter un Ouvrage tout neuf de fa ma-
-nufacture financière- de Surênes ; nous (avions déjà
-très bonne part que M. Ciaviere.Secrétaire de M.
-de Mirabeau , dans cette partie â avait des fouf-
-criptions toutes prêtes , pour quatre cents mil-
-lions en boucles d'argent, cuillierSj fourchettes,
-plats à barbe., Sec. Sec. Il eût été aifé d'hypo-
-téquer tous ces petits prêts fur les biens du
-Clergé y nous connaitTons même plufieurs per-
-fonnes qui fe fuilent contentées pour leurs
-boucles de col d'un privilège fpécial > exempt
-de toute retenue fur l'Abbaye ce Buzay.
-Quand les dix-neuf cents mille contrats ,
-grolTès, minutes , lignifications qu'aurait entraî-
-nés cette opération, fe feraient trouvés faits Se
-infinués ( i ) 3 alors on eût pu procéder fans
-
-(i) Cette befogne aurait le double avantage de pro-
-curer au Domaine des fomnies immenfes Se d'occuper cet
-hiver une foule de bras,, qui feront oififs faute d'emploi.
-Les Bureaux de la Compagnie , à vie , auroient préfenté
-un local tout près x & d'autant pius apte à la chefe
-qu'ils n'ont fervi à rien depuis la création de cet é:a-
-bliiTemenr moral; c'eût été une perte pour les Diirricts-,
-où les Commis de cette Compagnie fe font jettes à corps
-perdus , comme Prcûdcns ou comme Secrétaires , fauta
-d'occupatioa,
-
-A |

sources/actes-des-apotres/1789/11/actes-des-apotres-no-003.p6.txt

@@ -1,25 +0,0 @@
-(S)
-
-rifque à la création d'une caitfe de crédit. La
-proportion numérique du tiers au quart eût
-permis de faire pour quinze cents millions de
-papier , avec lequel on eût rembourfé tout l'ar-
-riéré , anticipé le paiement des anticipations
-non échues > fupprimé tous les impots défaf-
-treux & acquitté pendant deux ans toutes les
-charges publiques fans aucune contribution.
-
-Un tel plan , confîgné , a quelque chofe près ,
-dans le Courier de Provence, avait rempli de
-joie la Ville &: la Province , on n était fâché
-que du retard de fes auteurs à le propofer , afin
-de le voir en exécution immédiate; un moment
-a détruit ce beau rêve : M. Necker eft venu
-lefamedi, 14, à l'Afflmblée Nationale , faire
-ledture d'un plan provifoire de finance pour une
-année feulement; il compte, pour la perception
-de 175)0, fur la plupart des anciennes impo-
-rtions; c'eft encore fur de vieilles habitudes,
-fur de vieux fentimens qu'il fonde l'opération
-qui doit combler le déficit de 1789 & 175)0;
-il annonce que fes idées font fixées fur le plan

sources/actes-des-apotres/1789/11/actes-des-apotres-no-003.p7.txt

@@ -1,26 +0,0 @@
-( 7 )
-régénérateur & univerfel de nos finances ; mais
-qu'il lui paraît prudent) avant de propofer quel-
-que chofe à faire , d'attendre que l'AlTemblée
-ait ceiTé de défaire ; enfin 3 à la honte de l'ef-
-prit philofophifte & économise du tems , c'eft-
-cette odieufe CaiiTe d'Efcompte qui doit encore
-nous fournir tous ces premiers moyens de régé-
-nération, A la vérité, on ne lui demande que des
-fecours modérés; un appel de fonds de cinquante
-millions d'argent effectif, afin de pouvoir prêtée
-cent foixante-dix millions a la chofe publique &
-décupler les fecours qui manquent au commerce»
-aux manufactures & aux Créanciers de l'Etat :
-mais il eft aifé de voir que ce ne font encore
-que des palliatifs; lec moyens d'exécution éprou-
-veront les plus grandes difficultés. M. Necker -
-eft aimé, on le fait, mais le raifonnement , dans
-pareille occafion , ne doit il pas l'emporter fur
-le fentiment. Le maguétifme de M, Claviers
-le raifonneur eût attiré d'emblée , à fa caifîe
-quatre cents millons d'orfèvrerie; il eft impofiible
-que l'amour ôV le refpect dûs à la vertu de
-M. Necker procurent à la Banque Nationale les
-
-a4

sources/actes-des-apotres/1789/11/actes-des-apotres-no-003.p8.txt

@@ -1,27 +0,0 @@
-cinquante millions en écus qu'il lui faut pour
-réuflîr ; c'eft en vain qu'il répond d'avance à
-touresJes objections : c'tft en vain qu'il invoque
-le fentiment patriotique de< Français, la bien-
-veillance nationale; nous fomms arrivés à une
-époque où ces mots doivent nous être étrangers.
-Aux yeux de 1 homme qui réflfchit, le raient
-de M. Necker eft du p!u; grand danger Le Fran-
-çais, fur- tout, depuis q: elque tems , ne doit
-plus agit par inftinct , par impullion. Voyez
-comme M. de Mirabeau entraîne toutes les opi-
-nions _, par la (impie logique nue & décharnée.
-Chez lui la raifon eft tout , & le cœur n'y
-entre pour rien, Ces triomphes du fentiment
-n'ont d'ailleurs qu'une durée éphémère , ceux
-de la métaphyiîque pofent fur dts bafes im-
-muables j c'eft ainfi que tout eft vitalité > mais
-que tout eft deftruction dans la zone torride,
-& les gîaces majeftueufes qui couvrent le pôle
-ont commencé & finironr avec Je globe.
-
-Peuple français , ceflez donc de vous agiter
-d'erreurs en erreurs j préfervez-vous de toute
-atteinte du fentiment, concentrez vous , re-
-froid i (lez- vous > (î Ton peut s'exprimer ainfi ,
-
-A

sources/actes-des-apotres/1789/11/actes-des-apotres-no-003.p9.txt

@@ -1,26 +0,0 @@
-(9 V
-
-& que le Comte de Mirabeau vous dife par
-ma bouche qu'un Roi n'a plus befoin d être aimé
-C'eft un refte de fuperftftion qu'il faut détruire*
-Nous fommes arrivés à cet âge mûr où les
-pallions vives ne tyrannifent plus. Les Rois
-doivent aujourd'hui être envifagés comme de
-vieilles maîtreiTes qu'on a trop Iong-tems ga-
-rces. Un monarque n'elt que le préfixent du
-comité d'exécution. Toute fa (cisnce doit con-
-fifter a bien pofer une queftion. Chaque légif-
-lature lui votera des remercîmens ., & déformais
-l'hiftoire fécuîaire des monarchies doit fe com-
-pofer au plus de cinquante adrefTes du corps lé-
-gitîatif. Si la confticution nouvelle , comme
-on n'en peut pas douter , d'après l'influence
-falutaire du parti qui dirige l'aiTemblée , dépouille
-le Roi de toute autorité _, cette découverte pourra
-aller de pair avec celle de la poudre à canon.
-Le courago n'entre pour rien dans la compo-
-fîtion d'une armée } la nature ne doit entrer
-également pour rien dans la conft'nuion d'un
-grand empire. La règle & le compas décident
-les victo;res plus encore que les bayonnetres,
-c'eft avec une équerre & non par dçs montagnes