Skip to content
Catecharium

Published rights research for a work the Archive cites but does not reproduce.

Rights dossier — Lake, The Apostolic Fathers vol. II: Shepherd of Hermas, Martyrdom of Polycarp, Epistle to Diognetus

Researched: 2026-08-11, RIGHTS-DOSSIER drip (night 7). Ticket LS-318. One dossier clearing three works, as the acquisition list anticipated.

Verdict: CLEAR for all three. Pre-1930 publication, UK + US, established from the edition's own title page and printing statement read by eye off the page images.


(a) FIRST FINDING — the identifier named in the source list DOES NOT EXIST

docs/sourcing/study-bible-source-list.md and LS-318 both name IA apostolicfathers02lake (1913). That identifier returns an empty metadata record — there is no such item. Anything downstream that hard-codes it will 404, and the failure looks like a network problem rather than a bad id.

The correct item, located and verified this session, is apostolicfathers02lakeuoft — the University of Toronto scan, 414 images, _djvu.txt, _jp2.zip and _scandata.xml all present. (A second usable copy exists at TheApostolicFathersV2, a 1919 printing; not used here.) The source list and the ticket have been corrected.

IA's own possible-copyright-status = NOT_IN_COPYRIGHT field on that item was not relied on and is not evidence, per the standing rule. The basis below comes from the book.


(b) The edition — title page and printing statement, read off the page images

Leaf 0009 (the title page), read directly:

THE APOSTOLIC FATHERS WITH AN ENGLISH TRANSLATION BY KIRSOPP LAKE IN TWO VOLUMES II THE SHEPHERD OF HERMAS THE MARTYRDOM OF POLYCARP THE EPISTLE TO DIOGNETUS [Heinemann windmill device, W·H] LONDON : WILLIAM HEINEMANN NEW YORK : G. P. PUTNAM'S SONS MCMXVII

Toronto library stamp and accession H21563 / 5.4.44 in the margins.

Leaf 0010 (the verso), read directly:

First Printed, 1913. Reprinted, 1917.

That is the entire content of the copyright page — there is no copyright notice on it at all. The PD basis does not need that fact and does not rest on it; it is recorded because it is what the page says.

Note the US publisher differs from vol. I. Vol. I (1912) is Heinemann + The Macmillan Co.; this volume is Heinemann + G. P. Putnam's Sons. Loeb's American imprint changed between the two. Do not copy vol. I's provenance block across.

PD basis

Pre-1930 publication, simultaneously UK (Heinemann, London) and US (Putnam, New York); first printed 1913, this printing 1917.

  • United States. Everything published through 1930 is in the US public domain as of 1 January 2026. 1913 clears that by seventeen years; 1917 by thirteen. No notice or renewal question arises.
  • UK / EU (life + 70). Kirsopp Lake died 10 November 1946; +70 = expiry at the end of 2016. Public domain in the UK and EU as well, though the US basis is what governs us.
  • URAA. Restoration reaches only works still protected in the source country on 1 January 1996. This one was — Lake died 1946 — so the US term is the one that matters, and the US term is the pre-1930 rule above, which is unconditional for published works. (The same reasoning as vol. I; see rights-epistle-of-barnabas.md for the full statement.)
  • Scan. A faithful photographic reproduction of a two-dimensional public-domain work generates no new US copyright (Bridgeman v. Corel).

(c) Leaf-to-page map — ONE regime, unlike vol. I

Printed page = leaf index − 12, uniformly, printed pages 2 through 395.

From apostolicfathers02lakeuoft_scandata.xml (394 numbered leaves, a single offset regime beginning at leaf 14 → p. 2). Verified against actual page images at four widely-spaced points: leaf 0019 = p. 7, 0317 = p. 305, 0325 = p. 313, 0363 = p. 351.

This volume has none of vol. I's duplicate-leaf problem. A constant +12 is safe here. Cached: data/raw/rights-2026-08-11/scandata02.xml.


(d) The Shepherd of Hermas — pp. 1–305

Introduction pp. 1–6; bilingual text 6/7 – 305. By far the largest work in either volume (~300 pages, roughly the size of everything else in vol. II combined).

Fingerprints verified against page images:

  • Vision 1, i. 1 — printed p. 7, leaf 0019. Read directly: opening recto headed THE SHEPHERD / VISION 1 / I, then

    "1. HE who brought me up sold me to a certain Rhoda at Rome. After many years I made her acquaintance again, and began to love her as a sister."

    Marginal glosses "Hermas and Rhoda", "Hermas goes to Cumae"; page number "7" at the foot.

  • Similitude X. iv. 5, the close of the work — printed p. 305, leaf 0317. Read directly: running head THE SHEPHERD, SIM. X. IV. 2–5,

    "5. Now after he had spoken this he rose from the couch, and took the shepherd and the maidens and departed, but said to me that he would send back the shepherd and the maidens to my house."

    Foot of page: "305", VOL. II. and signature X.

THE ADDRESS PROBLEM — Hermas does not fit the vol. I converter's shape

Every work cleared so far in this corpus is chapter:verse. Hermas is not. Lake prints a three-part division — Visions, Mandates, Similitudes — and the running heads show the address is four tiers with an optional third:

| Printed running head | Address | | :-- | :-- | | THE SHEPHERD, VIS. I. i. 3–8 | Vision 1, chapter i, vv. 3–8 | | THE SHEPHERD, VIS. V. 1–5 | Vision 5, no chapter tier, vv. 1–5 | | THE SHEPHERD, MAND. II. 2–7 | Mandate 2, no chapter tier, vv. 2–7 | | THE SHEPHERD, MAND. IV. i. 3–8 | Mandate 4, chapter i, vv. 3–8 | | THE SHEPHERD, SIM. V. ii. 5–11 | Similitude 5, chapter ii, vv. 5–11 | | THE SHEPHERD, SIM. IX. xxxiii. 3 | Similitude 9, chapter xxxiii (Sim. IX runs to xxxiii) |

So: book-division (Vis/Mand/Sim) → number (upper roman) → chapter (lower roman, PRESENT FOR SOME UNITS AND ABSENT FOR OTHERS) → verse (arabic). Vision 5 and most Mandates have no chapter tier; Vision 1–4 and the Similitudes do. Similitude IX alone runs to thirty-three chapters.

That is an address-design decision, and it should be made before any transcription starts, not discovered mid-build. It is a bigger question than the letter tier Ignatius needs, and it is the reason this work is NOT a drip-night build. The two smaller works in this volume are.

Structural facts that will matter to that decision: the divisions are printed as VISION 1 / MANDATE 2 / SIMILITUDE 5 on their own lines in the English text, with the chapter (I, II, …) as a separate centred line where present; verse numbering restarts at 1 within each addressable unit.


(e) The Martyrdom of Polycarp — pp. 307–345

Section half-title after p. 305; introduction pp. 307–311; bilingual text 312/313 – 345. Twenty-two chapters, I–XXII, plus an unnumbered salutation, plus an appended alternative ending.

Fingerprint verified against the page image — printed p. 313, leaf 0325. Read directly: opening recto headed THE MARTYRDOM OF ST. POLY- / CARP, BISHOP OF SMYRNA, then the unnumbered salutation:

"THE Church of God which sojourns in Smyrna, to the Church of God which sojourns in Philomelium, and to all the sojournings of the Holy Catholic Church in every place. 'Mercy, peace and love' of God the Father, and our Lord Jesus Christ be multiplied."

then chapter I:

"1. WE write to you, brethren, the story of the martyrs and of the blessed Polycarp, who put an end to the persecution by his martyrdom as though adding the seal."

Marginal gloss "Introduction"; page number "313" at the foot.

Two structural traps

  1. The salutation is unnumbered, exactly as in Polycarp's own Epistle (see rights-2-clement-and-polycarp.md §d). Same address decision, same rule: do not silently renumber it 1:1.

  2. After chapter XXII.3 the edition prints a SECOND ending with its own numbering. Printed p. 343, centred head:

    ANOTHER CONCLUSION FROM THE MOSCOW MANUSCRIPT.

    and beneath it sections numbered 2, 3, 4, 5, 6 — running heads read MARTYRDOM OF POLYCARP, 3–6 with no chapter numeral at all, in contrast to every other recto in the work. It runs to p. 345 and ends "…for ever and ever, Amen."

    A converter that keys on "arabic number at line start = verse of the current chapter" will append the Moscow epilogue's 2–6 to chapter XXII, silently producing a chapter XXII with duplicate verse numbers and a doubled doxology. It needs its own unit, addressed as an appendix rather than as a chapter.


(f) The Epistle to Diognetus — pp. 348–379

Section half-title after p. 345; introduction pp. 348–350; bilingual text 350/351 – 379. Twelve chapters, I–XII. The smallest and simplest of the three.

Fingerprint verified against the page image — printed p. 351, leaf 0363. Read directly: opening recto headed THE EPISTLE TO DIOGNETUS, chapter I:

"SINCE I perceive, most excellent Diognetus, that you are exceedingly zealous to learn the religion of the Christians…"

then chapter II:

"1. COME then, clear yourself of all the prejudice which occupies your mind, and throw aside the custom which deceives you, and become as it were anew man from the beginning…"

Marginal glosses "Introduction", "Discussion of the Gods of the heathen"; page number "351" at the foot. The work closes on p. 379 at XII.9, "…and through it the Father is glorified; to whom be glory for ever, Amen."

Trap: chapter I carries NO verse number. Lake prints it as a single unnumbered paragraph; numbering begins at II.1. Editions and citations commonly address that paragraph as Diognetus 1:1, so an ingest that faithfully records "chapter I, no verse" will not match an incoming citation of 1:1, and an ingest that invents "1" is silently editing the edition. Note which was chosen in the provenance rather than leaving it implicit. Chapters XI–XII are widely held to be a later appendix by a different hand; Lake prints them continuously with no division, so nothing in the build needs to act on that, but the reader-facing note should say it.


(g) Common facts for a converter

  • Facing-page Greek (verso) / English (recto) throughout, the same layout transcribe-barnabas.ts handles. No Latin-verso trap here — that is a vol. I / Polycarp problem only.
  • Verso running head: THE APOSTOLIC FATHERS. Recto: <WORK>, <range>. Opening rectos are headed with the full title instead of a range, as in vol. I.
  • Marginal summary glosses down the outer margin of the English rectos in all three works, same as vol. I.
  • Greek characters appear in English-recto FOOTNOTES (e.g. Hermas p. 7 n.1 quotes ἀνεγνωρισάμην; Mart. Poly. p. 313 n.1 quotes μαρτύριον / μαρτυρία). The existing rule holds and matters here: on a recto, strip Greek characters, never drop the line.
  • Leaf offset a constant +12; scandata02.xml is available if a map is preferred.

(h) Verdict and build gates

Verdict: CLEAR — all three works, one rights basis.

Suggested slugs: shepherd-of-hermas, martyrdom-of-polycarp, epistle-to-diognetus.

Sequencing recommendation: build Diognetus and the Martyrdom as ordinary drip-night items; hold the Shepherd of Hermas for a deliberate address-design decision first. Diognetus is 28 pages of straight chapter:verse; the Martyrdom is 32 pages with two named quirks. Hermas is 300 pages with a novel four-tier address whose third tier is optional, and getting that wrong is expensive to undo.

Gates for all three:

  1. Use IA identifier apostolicfathers02lakeuoft. apostolicfathers02lake does not exist.
  2. English rectos only.
  3. Template transcribe-barnabas.ts, head strings and chapter ranges adapted; leaf offset +12.
  4. Corrections-as-data for OCR damage, in the TEXT_CORRECTIONS style.
  5. Work-specific: Martyrdom — unnumbered salutation, and the Moscow-manuscript conclusion (pp. 343–345, sections 2–6) must be its own unit, not chapter XXII's tail. Diognetus — chapter I has no printed verse number; record the decision. Hermas — address design first; assert Vis. I–V, Mand. I–XII, Sim. I–X, with Sim. IX running to chapter xxxiii.
  6. Assert the fingerprinted wording at Hermas Vis. I.i.1 and Sim. X.iv.5; Mart. Poly. salutation and I.1; Diognetus I and II.1.
  7. Provenance: Heinemann / G. P. Putnam's Sons, first printed 1913, this printing 1917 — not Macmillan, and not 1912.

Sources

  • https://archive.org/details/apostolicfathers02lakeuoft — the dated scan (title page and printing statement read by eye, leaves 0009 and 0010)
  • data/raw/rights-2026-08-11/lake02_djvu.txt — full volume OCR (gitignored)
  • data/raw/rights-2026-08-11/scandata02.xml — IA leaf→printed-page map (gitignored)
  • data/raw/rights-2026-08-11/v2_leaf*.jpg — cached page images (gitignored)
  • docs/sourcing/rights-epistle-of-barnabas.md — the parallel vol. I basis and its full US-term reasoning

2026-10-01 (research sprint): THE SHEPHERD OF HERMAS, A CONCRETE ADDRESS PROPOSAL

Design proposal only; nothing implemented, no database touched. Rights are CLEAR and unchanged. This confirms the 2026-10-01 content-night decision (three division-works, chapter tier declared per unit) and supplies what that section left open: the exact label grammar, the worked examples, the collision analysis, the alignment rows, and a precise statement of what is and is not inside the existing schema. Evidence read in the cached OCR of the volume, rebuilt as plain text in apps/api/data/raw/research-2026-10-01/lake2.txt (from lake-vol2-2026-09-28/lake02_djvu.xml; leaf numbers below are that file's LEAF n markers, not IA n numbers).

1. How the corpus's addressing actually works (checked, not assumed)

  • A compiled work is Work -> unit (number, title) -> chapter (number, title) -> section (n, label), render/<unit>.json. Numbers must be positive integers (lwm.ts, front-matter check); the label is a free string.
  • The default heading is two-tier ("unit, section"): second-epistle-of-clement is unit = chapter, sections = verses, labels "1, 1", "1, 2".
  • Three-tier works already exist: adversus-haereses-st-irenaeus-of-lyons carries citationScheme: book-chapter-section and gets headings "5, 12, 1" through the opt-in sectionHeadingFromLead plus stripSectionLead (the label is a bold lead on each section, hidden from the reader); confessions is book.chapter.section. citationScheme is a free string with no enum (lwm.ts), unitNoun exists for "vision / mandate / similitude", and unitCountLabel exists for mixed units.
  • A DS stub is a node whose slug is ds-<heading>; where the stub's locus is coarser or finer than the transcription, a row in a *-ds-alignment.csv maps it (work_slug, ds_heading, ds_stub_node_slug, new_unit_number, new_unit_title, new_chapter, new_section_range, confidence, note). The loader (build-ds-reader-targets.ts) reads EVERY *-ds-alignment.csv in the directory generically, so a new Hermas file needs no code.
  • What the Hermas stubs and citations are (catechism-sources.json, the only places Hermas is cited; the manifest has no other Hermas row): visiones-pastoris-hermas heading 2, 4, 1 edge (CCC 760); mandata-pastoris-hermas heading 2, 1, 1 edge in the citation data (CCC 2517; the registry says 2 and I did not reconcile the difference). There is no Similitudes row at all. Both stubs are "number, number" in the CCC's own arabic Funk/Gebhardt-Harnack style: division is carried by the work's name, never by the address.

2. Lake's address, as measured on the pages

THE SHEPHERD, vis. ii. iv. 2-iii. i. 2 (running head, leaf 34): division, upper-roman number, lower-roman chapter, verse. Chapter numerals print on their own line before the "1." (Vision II chapter iv opens "IV / 1. And a revelation was made to me, brethren, while I slept ..."). Chapterless units print no numeral at all: Mandate 2 prints "MANDATE 2" then "1. He said to me: Have simplicity and be innocent ..." straight into verses 1 to 7 (leaves 80-81). Per the earlier measurement the chapter tier is a per-number datum (Visions 1 present, 5 absent; Mandates 5, 10, 12 present, 2, 9, 11 absent; Similitudes 5, 6, 7, 9, 10 present, Similitude I absent).

3. THE PROPOSAL

Three works, division-as-work, unchanged from the earlier decision: visiones-pastoris-hermas, mandata-pastoris-hermas, similitudines-pastoris-hermas. Inside each, with unitNoun of vision, mandate, similitude:

| Tier | Lake's | Our node | Stored as | | :-- | :-- | :-- | :-- | | unit | upper-roman number (Vision I) | unit, number N | integer; title "Vision I" etc. | | chapter | lower-roman (i, ii ... xxxiii), when printed | chapter, number c | integer; title "Chapter i" etc. | | chapterless unit | nothing printed | ONE anonymous chapter numbered 1 | the Clement precedent (a unit with a single chapter); its title is never shown | | section | verse | section, n | integer |

The label (the heading, the address) is arabic and its arity is a function of the unit:

  • unit with chapters: N, c, v
  • chapterless unit: N, v

with citationScheme: division-chapter-verse (free string), sectionHeadingFromLead: true, stripSectionLead: true, exactly the Irenaeus mechanism. The reader sees Lake's own numerals (unit and chapter displayed in roman in the title strings, e.g. "Vision I, chapter i"), while the machine address stays arabic to match the CCC-derived stubs and every other work in the archive.

The three examples

| Lake's citation | Work | unit | chapter | verse | label | text it lands on (checked) | | :-- | :-- | --: | --: | --: | :-- | :-- | | Vision I.i.3 | visiones-pastoris-hermas | 1 | 1 | 3 | 1, 1, 3 | "After some time, while I was going to Cumae, and glorifying the creation of God ... I became sleepy" (leaf 16) | | Mandate II.2 | mandata-pastoris-hermas | 2 | (none, anonymous chapter 1) | 2 | 2, 2 | "In the first place, speak evil of no one, and do not listen gladly to him who speaks evil" (leaf 80) | | Similitude IX.xxxiii.x | similitudines-pastoris-hermas | 9 | 33 | x | 9, 33, x | the example needs a correction: Similitude IX chapter xxxiii has only verses 1 to 3 in Lake (leaf 306; running head "sim. ix. xxxii. 5-xxxiii. 3"). The last address is 9, 33, 3 ("Listen now, said he, about them. These are those who heard my commandments ..."). 9, 33, 10 does not exist, and the build must assert the verse count so it cannot be produced |

Also noted while there: Similitude X is printed in Latin only on the facing page and Lake's English runs on the odd pages; chapter xxxiii's OCR numeral reads "XXXITI" and an earlier one "XXXII" three times, so the chapter number must come from the declared sequence and be asserted against the printed numeral, never read straight from the OCR numeral, which is the discipline expectedChapters already enforces for the other Lake works.

4. The two stubs against this design

  • mandata-pastoris-hermas 2, 1 resolves EXACTLY, because Mandate II is chapterless and its labels are 2, v. It lands on "Have simplicity and be innocent and you shall be as the children who do not know the wickedness that destroys the life of men" (Mandate II.1, leaf 80). No alignment row.
  • visiones-pastoris-hermas 2, 4 cannot match exactly, because every Vision II label has three parts (2, 4, v). It needs one alignment row. The CCC's locus is Vision II, chapter 4, and the text the CCC quotes ("she was created the first of all things ... for her sake was the world established") is Vision II.iv.1, so the row can point at section 1 with high confidence:
work_slug,ds_heading,ds_stub_node_slug,new_unit_number,new_unit_title,new_chapter,new_section_range,confidence,note
visiones-pastoris-hermas,"2, 4",/archive/works/visiones-pastoris-hermas/ds-2-4,2,Vision II,4,§1,high,"Chapter-granular CCC locus (Vis. 2, 4); lands on II.iv.1 ... content-asserted at build time on 'created the first of all things'"

(the column layout above is the existing file's; the unit-title and note wording is a sketch). The assertion in the build is "section 2, 4, 1 contains 'created the first of all things'", the Benedict pattern. The two stubs are both "two numbers" but mean different things (Mandate II + verse against Vision II + chapter), which is the one mistake that strands an edge; the label rule above is what makes that visible in the data.

5. Collision analysis

Within one work, can a two-part and a three-part label be confused? No, because arity is a function of the unit and the unit is the first number. In visiones, units 1 to 4 are chaptered and unit 5 is not, so 5, 4 can only mean Vision V verse 4 and 2, 4 can only mean Vision II chapter 4; there is no 5, 4, v and no 2, v. This holds only while the build declares the per-unit chapter table and asserts it (a stray chapter numeral OCR'd in a chapterless unit would otherwise create a 5, 1, 3 beside the 5, 3 family silently). The declared inventory is therefore a required build input, not a convenience. The earlier section already found the head OCR too lossy to produce it (2 of 5 Visions, 6 of 12 Mandates resolved), so it must be read off the pages.

Alternatives considered and rejected:

  • Pad chapterless units with a fake chapter in the label (2, 1, 2): breaks the Mandate II stub's exact match, which is the cheaper of the two stubs and the one with the most edges.
  • Flatten chapters into a continuous verse run per unit: invents numbers Lake does not print and loses the printed address, the opposite of the ADDRESS-DEPTH RULING.
  • One work with four tiers (division as a tier): pushes the division into the address, collides with how the CCC cites (division named by the work), and would need a schema change for no benefit.
  • Use the Gebhardt-Harnack continuous numbering (Vis 1-25, Mand 26-49, Sim 50-114): it is not what Lake prints, and not what our stubs use.

6. Schema and code: what is and is not a change

Within the existing schema, no change: Work, WorkNode, lwm.ts, the render JSON, unitNoun, citationScheme (free string), the Irenaeus lead-as-heading mechanism for three-part labels, and the alignment CSV and its generic consumer all already do this. No Prisma migration, no lwm.ts change.

One converter change, unchanged from the earlier section but now fully specified: LakeWorkSpec (transcribe-apostolic-fathers.ts, line 79) is strictly chapter:verse (expectedChapters, fingerprints: { c, v, contains }, stopAt(chapter, verse, text)). It needs (a) an OPTIONAL third tier with a per-unit declaration of whether the tier exists and how many chapters and verses each carries (asserted), (b) a unit-number tier above chapter (a Hermas unit is a division number, not a chapter), (c) fingerprints { unit, c?, v, contains }, and (d) emission of the two label shapes above. All fourteen published Lake works must still build unchanged; run the LS-432 scratch-build diff across the corpus before and after, which is the guard the earlier section named.

7. Registry annotation (verbatim, for the supervisor to apply)

2026-10-01 RESEARCH SPRINT: HERMAS ADDRESS PROPOSAL (design only, WITHIN the existing schema, one converter change). Three division-works as already decided. Inside each: unit = division number (unitNoun vision / mandate / similitude), chapter = Lake's lower-roman tier when printed, section = verse; chapterless units carry one anonymous chapter. Labels are arabic and their arity is a function of the unit: N, c, v for chaptered units, N, v for chapterless ones, emitted through sectionHeadingFromLead plus stripSectionLead (the Irenaeus mechanism; citationScheme: division-chapter-verse is a free string). Examples: Vision I.i.3 is 1, 1, 3, Mandate II.2 is 2, 2, Similitude IX.xxxiii is 9, 33, v and Lake prints only verses 1 to 3 there, so the last address is 9, 33, 3 and 9, 33, 10 does not exist (assert verse counts). Stub mandata 2, 1 resolves EXACTLY (chapterless Mandate II, verse 1); stub visiones 2, 4 needs ONE alignment row (hermas-ds-alignment.csv, Vision II chapter 4 section 1, the CCC passage "created the first of all things" is II.iv.1, content-asserted). No Prisma or lwm.ts change; the alignment CSV is read generically. The one code change is LakeWorkSpec: optional third tier with a per-unit declared and asserted chapter inventory (which MUST be pinned off the pages, the head OCR resolves only part of it), a unit tier above chapter, {unit, c?, v} fingerprints and two label shapes; the LS-432 whole-corpus scratch-build diff is the guard. Chapter numbers come from the declared sequence, asserted against the printed numeral, never read from OCR numerals (XXXITI, XXXII x3 seen). Dossier: rights-lake-vol-ii.md.

Question for the founder

None required for the address design. One small question hangs on it: the registry's own build order (Mandates, Visions, Similitudes, with Similitude IX its own night) stands. Recommendation: build Mandates first (smallest, carries the most edges, its stub needs no alignment row), as already decided.

© 2026 Catecharium