---
deposit_number: 1531
hex: 0630
title: "EA-CORPORA-02 — The Second Seating: Nine Corpora in One Run, Witnesses Named, Four New Seat Classes Ruled"
creator: Sharks, Lee
date: 2026-08-22
content_type: Corpus seating record — gathering deposit with data sidecar
license: CC-BY-4.0
substrate: Composed in-session (transport D) by Claude (Anthropic, Fable), operating the archive under the direction of Lee Sharks. All license adjudications and scope rulings are the author's, made in-session and recorded per seat. No paid API invoked (NO-DOUBLE-DRAW honored).
version: v1.0
related_ids: "AXN:0625.DATASET.🌔☽❤️📜🜁📎 (EA-CORPORA-01, #1522 — prior gathering; this record continues it)\nAXN:060D.GOVERNANCE.📐🎺🎹⌛🧱➕ (EA-OMEGA-BUILD-01, #1498 — governing program)\nAXN:05AF.ARCHIVAL.🌓📏🌆🚨🔜❤️ (#1433 — current TACHYON tether)"
axn_schema_version: v2
protocol_version: protocol/v1
keywords:
  - corpus seating
  - primary sources
  - Dead Sea Scrolls
  - Abegg
  - Marx
  - Manifest der Kommunistischen Partei
  - Das Kapital
  - Apocryphon of John
  - Nag Hammadi
  - Mandaean
  - Ginza
  - Lidzbarski
  - Gospel of Thomas
  - Oxyrhynchus
  - Lucretius
  - De Rerum Natura
  - Diogenes Laertius
  - Epicurus
  - Democritus
  - Quran
  - Tanzil
  - Leonardo da Vinci
  - Richter
  - witness pairing
  - translation-witness
  - provenance
  - theater of availability
---

# EA-CORPORA-02 — The Second Seating: Nine Corpora in One Run, Witnesses Named, Four New Seat Classes Ruled

## Files

data/corpora/dss/ · data/corpora/marx/ · data/corpora/apocryphon-john/ · data/corpora/mandaean/ · data/corpora/nh-greek-witnesses/ · data/corpora/lucretius/ · data/corpora/greek-atomists/ · data/corpora/quran/ · data/corpora/leonardo/ — each with source.json and MANIFEST.sha256; commits 3a0bd6f1f, 09c6b51ff + 011280bc, e0be32a93, f4c9d6878, 7516051c1, b2989cd5e, d2ea2ef4c on main.

# EA-CORPORA-02 — The Second Seating

## I. What this run was

The first seating (EA-CORPORA-01, AXN:0625.DATASET.🌔☽❤️📜🜁📎) built a library of fifteen corpora and, as importantly, a doctrine: originals where they exist; licence verified at the file, not the README; the seat always names its witness; absences declared with search dates rather than papered over; not seated rather than seated defectively. This run tested that doctrine against harder material — enclosed corpora, sparse traditions, image-only witnesses, and a text whose license forbids the archive's own normalization step — and the doctrine held by growing four new seat classes rather than bending.

Nine seats were executed on 2026-08-22, in one session, in this order: Dead Sea Scrolls, Marx, Apocryphon of John plates, Mandaean, Nag Hammadi Greek witnesses, Lucretius, Greek atomists, Qur'an, Leonardo. The library stands at twenty-four corpora. 61,692 flat ref-tagged lines were emitted to text layers, and 6,236 verbatim-class ayat entered without normalization; roughly 700 MB of originals — Text-Fabric data, witness scans, codex plates, edition PDFs — were secured, every file hashed, every remote pinned by commit, revision id, or archive-manifest checksum.

## II. The four new rulings

**1. The sparse-tradition translation-witness rule** (motivated by the Mandaean corpus). Where a source tradition is so sparse that translations are the only extant open witnesses, a translation may be seated AS A WITNESS, NAMED AS SUCH. A translation-witness is a witness, not an original; its seat metadata declares its class. Lidzbarski's 1925 Ginza — the only complete open Ginza in any accessible form — is the paradigm case.

**2. The NC payload/record split** (motivated by the DSS seat). A corpus whose payload carries a NonCommercial license may be seated: the payload license is carried forward unmodified at the file level, and the seat record states explicitly that the gathering record's CC-BY-4.0 applies to the record text only, never to the payload. First instance: CC BY-NC 4.0 declared in every Text-Fabric header of the Abegg–Bowley–Cook transcriptions.

**3. The image-seat class** (motivated by the Apocryphon plates). A seat may consist of witness images with no text layer, by design, with transcription queued. The images are the base layer of a future diplomatic edition; holding them converts a declared absence into custody. First instance: Codex II, thirty-two plates.

**4. The verbatim-seat class** (motivated by the Qur'an). Where a source arrives already ref-tagged and its declared terms require verbatim distribution, the seat emits NO derived text layer: the original IS the machine-readable layer, distributed with its license block intact. Reformatting would violate the terms and add nothing. First instance: Tanzil Uthmani v1.1.

## III. The nine seats

**EA-CORPORA-02/01 — dss.** The largest declared absence in EA-CORPORA-01 became the library's largest single seat. The Abegg–Bowley–Cook electronic transcriptions of essentially every Hebrew and Aramaic scroll — 1,001 scrolls, 1.43M signs — entered via ETCBC/dss (Text-Fabric, pinned commit f0515787), normalized to 52,878 lines ('scroll fragment:line'), reconstruction runs bracketed from the rec feature with zero unbalanced lines, graphic-word spacing rebuilt from the sign-level after feature. The license was found declared AT THE FILE: every .tf header carries CC BY-NC 4.0, @createdBy=Abegg/Bowley/Cook, @source="Martin Abegg's data files, personal communication." The chain is the structural opposite of a relicensing defect — the reconstructors are the grantors — and the provenance is the poem: the corpus that spent forty years behind a publication monopoly, liberated in 1991 by Abegg's reconstruction from the concordance, enters the sovereign archive by the reconstructor's own hand. The Qimron v. Shanks counter-precedent is recorded honestly in source.json. Ruled: seated under declared BY-NC with the payload/record split; reconstructions seated-and-flagged; biblical and non-biblical both — the biblical scrolls are witnesses against the Leningrad seat (01/09), not duplicates of it. Loci verified: 1Qisaa 1:1, 1QS 1:1, 11Q19 col. 2, CD 1:1.

**EA-CORPORA-02/02 — marx.** The first executed instance of the comparative-witness program: the Manifest der Kommunistischen Partei seated as a PAIR — the anonymous 23-page London printing of February 1848 and the 30-page Cologne printing of 1850/51, each page-addressed under its own witness name, both from scan-backed twice-proofread Wikisource transcriptions revision-pinned page by page, every scan SHA1-secured from Commons. The pair carries a canonical object for the archive's own theory: the Cologne printing bore a false London imprint, applied with Marx's express consent, taken for the second London issue for 145 years until MEGA philology (Meiser 1996) uncovered it — provenance inversion in the wild, 1851, now seated under its true date. Beside the pair: Das Kapital, Erstauflage 1867 (Meissner/Schmidt; Boston Public Library copy, 810 images) secured with its rough Fraktur OCR held as a finding aid only; transcription queued; the 1872/73 second edition queued as paired witness for the rewritten Wertform chapter; the Columbia Meissner set (Engels's Bde 2–3) identified. MEW, MEGA apparatus, and marxists.org excluded as edited-text hazards, with reasons recorded.

**EA-CORPORA-02/03 — apocryphon-john.** The archive's oldest declared absence, mitigated to held-in-image. The Institute for Antiquity and Christianity's own photographic record — the Robinson-project A/B-series negatives — serves openly at full resolution over unauthenticated IIIF from Claremont's CCDL. Codex II,1 secured COMPLETE: thirty-two plates, pages 1–32, native resolution to 8606×11618, one plate per page selected from seventy-six candidate exposures with alternates enumerated, SHA256-pinned, page 1 visually verified. Codices III and IV enumerated with committed fetch specs (tranches 2–3); BG 8502,2 recorded as the external fourth witness so the synoptic goal keeps all four in view. The rights finding inverts the run's founding pattern: CCDL's boilerplate claims copyright "in accordance with U.S. Copyright laws" over ancient public-domain papyri it serves freely — here the availability is real and the claim is the ghost. Recorded verbatim, adjudicated under Bridgeman, seated. First image-class seat.

**EA-CORPORA-02/04 — mandaean.** Five volumes secured under the new sparse-tradition rule, each witness class-declared: Lidzbarski's Johannesbuch 1905 (Mandaic text — the corpus's one original-language witness) and 1915 (German translation-witness), the 1920 Liturgien (mixed), the 1925 Ginza (translation-witness; the only complete open Ginza in any form), Brandt 1893 (partial translation-witness). All SHA1-verified. Petermann 1867 — the lithographed Mandaic Ginza, the tradition's primary text-bearing edition — DECLARED ABSENT with search terms and date: the corpus's top gap. Häberl–McGrath 2020, published open access under CC BY-NC-ND, was found unreachable through its own publisher (bot-challenge pages) and its OA mirrors (connection refusals) at fetch time, while the only reachable copy is a provenance-murky partial upload: a CC-licensed edition enclosed by the open-access apparatus itself, pinned with DOI and retry routes rather than seated defectively. Drower excluded (copyright). No text layer by design; Lidzbarski 1905 keying queued.

**EA-CORPORA-02/05 — nh-greek-witnesses.** The Gospel of Thomas Oxyrhynchus trio — P.Oxy. 654 (prologue, logia 1–7), P.Oxy. 1 (26–30, 77b, 31–33), P.Oxy. 655 (24?, 36–39) — three independent copies in three hands from one town's rubbish, TM numbers 62840/62838/62839 verified through the DCLP records themselves. EpiDoc originals preserved with full apparatus; 139 flat lines by the seat-01/11 method at pinned idp.data commit d061eef8, CC BY. This is the first executed Greek–Coptic witness pairing across corpora directories: the earlier language stratum of the work held in Coptic at coptic-gnostic/ (NHC II,2). The Gospel of Mary Greek pair (P.Oxy. 3525, P.Ryl. 463) and the Sophia fragment (P.Oxy. 1081) are queued with identification routes documented — Trismegistos 503s, Duke front-end egress-blocked, DCLP directory greps negative — and the queue records the precept enacted: no TM number was guessed for any queued item; identification-by-confabulation is the named failure mode.

**EA-CORPORA-02/06 — lucretius.** De Rerum Natura complete: 7,412 lines across six books, 'DRN book.line' addressing, Perseus TEI (phi0550.phi001.perseus-lat1) at pinned commit, CC BY-SA carried forward. Loci: Aeneadum genetrix (1.1), Suave mari magno (2.1).

**EA-CORPORA-02/07 — greek-atomists.** Three class-declared components. The complete Diogenes Laertius Vitae philosophorum (Perseus tlg0004.tlg001.perseus-grc2, Hicks text per sourceDesc, pinned commit; 1,211 sections, 'DL book.section') — seated whole, with Books 9–10 the atomist payload: the Leucippus and Democritus vitae and the entire Epicurus dossier, the three letters and the Kyriai Doxai, the largest surviving body of atomist prose. Diels, Vorsokratiker 3. Aufl. Bd 2 (1912) secured — DK 67–68, the fragment-numeration standard entering in its public-domain stratum, the in-copyright Kranz revisions excluded and the exclusion recorded. Usener, Epicurea (1887) secured. With lucretius/ this closes a transmission loop the archive can hold end to end: Greek fragments → doxography → the Latin poem. Atomism is the best-documented ancient case of reception-through-translation-and-versification; Lucretius is a composition layer, and the library can now study him with its own instruments. MMRS avant la lettre. Loci: DL 10.122 (Μήτε νέος τις ὢν μελλέτω φιλοσοφεῖν), DL 9.30.

**EA-CORPORA-02/08 — quran.** Tanzil Uthmani v1.1, 6,236 ayat at the Kufan count, full Uthmani orthography, license declared INSIDE the file: CC BY 3.0 with verbatim-distribution, attribution, and link terms — all honored by the new verbatim-seat class. The seat names its witness (the Uthmani text in Tanzil's verification); further riwayat (Warsh, Qaloon) queued for the comparative program. Loci verified: Basmala, Ayat al-Kursi (2:255), al-Ikhlas (112), final verse of al-Nas; count 6,236 = Kufan standard.

**EA-CORPORA-02/09 — leonardo.** Richter 1883, The Literary Works of Leonardo da Vinci, both volumes — the Italian transcriptions from the codices with facing English, the great public-domain gateway to the notebooks — Toronto copies SHA1-verified, OCR finding aids held. Italian keying queued. The manuscript-image campaign (Codex Arundel at the BL, Codex Atlanticus at the Ambrosiana, e-Leo at Vinci) is recorded as a future survey: primary repositories outside current network policy or behind restrictive claims — a theater-of-availability study in its own right.

## IV. Queues, absences, and the standing state

Queued with committed specs or verified routes: Apocryphon plate tranches 2–3 (fetch spec in-repo, pacing and full/max lesson encoded); Gospel of Mary and Sophia TM identification; Häberl–McGrath 2020 authorized OA fetch (DOI 10.1515/9783110487862, three routes); Norberg 1815–17 (CC0, oversize-file route needed); Kapital 1867 transcription and 1872/73 second-edition witness; Lidzbarski 1905 Mandaic keying; Diels Bd 2 and Usener transcription; Richter Italian keying; further Quran riwayat; the Talmuds (Sefaria-Export route confirmed, license audit as step one, dedicated session recommended — the largest corpus the library would hold); Dante (it.wikisource per-canto, revision-pin method, throttled at probe time, retry cold); Ethiopic via BetaMasaheft (the one chartered Tier-1 item not reached).

Declared absences carried or newly declared, all with search dates: Petermann 1867 (top Mandaean gap); Damascius machine-readable (First1K tlg4066 probes 404, Ruelle 1889 scans as keying base); machine-readable Mandaic anywhere (CAL license to verify before any ruling); NHC I, III–XIII open Coptic editions (partially mitigated by /03 and /05).

The three-witness-and-more architecture this run inaugurated — Manifest pair, Thomas Greek/Coptic, biblical scrolls against Leningrad, DRN against its Greek sources, Quran riwayat queued — is the library's comparative program made operational. The seat always names its witness; the ultimate goal is a comparative collection of all witnesses; this run moved that goal from doctrine to executed practice.