Governing rules

• Keep every papyrus word in manuscript order and align only by a contiguous forward scan.
• Show every siglum that actually attests the displayed location; multiple agreeing papyri may share a compact cell.
• Preserve disagreeing readings separately in the data even when the interface later offers a compact view.
• When attesting papyri disagree, display the reading of the papyrus ranked first by the public papyrus chronology: earliest starting year in the paleographic date range, then lower Gregory-Aland papyrus number as the deterministic display tie-breaker.
• Attach compact-view sigla only to papyri that attest the selected displayed reading; retain dissenting papyri and readings in provenance.
• Use lost status for non-extant material; a leading loss requires INTF confirmation.
• Strip diacritics and expand nomina sacra for comparison only, never as an unrecorded alteration of stored evidence.
• Coverage stubs may use TAGNT only as a visibly identified provisional reconstruction, never as transcribed papyrus text.

Not permitted

• Silent TAGNT substitution
• Automatic correction of genuine variants
• Reordering manuscript words
• Treating verse coverage as proof that every word survives

Next certification action: Add per-reading provenance and a visible stub/reconstruction state, then finish CNTR-to-INTF word-level collation.

Governing rules

• Preserve every Sahidica word-group exactly and attach edition, source file, verse, occurrence number, diplomatic form, and SHA-256 provenance.
• Preserve Sahidica source order and word-group boundaries independently; never reshape the Coptic sequence merely to fit another tradition’s grid.
• Represent many-to-many correspondence with alignment links or spans while retaining provenance for every Sahidica source unit.
• Treat John 8 as the second logical chapter embedded in the distributed 43_John_07.tt file; keep John 7:53 and John 8:1–11 explicitly omitted.
• Describe SCRIPTORIUM segmentation and linguistic annotation as automatic source layers, not manually Coptologist-validated facts.
• Display Crum/KELLIA only as lexical aid.
• Admit Horner English verbatim only when Horner’s underlying Coptic is exact or nonlexically equivalent to the corresponding Sahidica span.

Not permitted

• Calling Sahidica simply Horner
• Claiming that one normalized edition represents every Sahidic manuscript
• Presenting automatic SCRIPTORIUM annotations as manually validated
• Using TAGNT, another tradition, a dictionary, or project-generated wording as Sahidic translation
• Filling Sahidica omissions from another tradition
• Extracting Logos content for public redistribution without written permission

Next certification action: Acquire a qualified Horner Coptic-and-English transcription for contextual translation units, while preserving the complete multi-source decision ledger; repair parallel placement separately.

Governing rules

• Display GA 03, not a critical edition proxy.
• Normalize case and accents only under the declared display policy.
• Preserve lacunae, supplied text, uncertainty, corrections, and selected scribal hand.
• Use TAGNT contextual English only after deterministic alignment to the INTF-controlled GA 03 word; verify its lexical identity against TBESG/Abbott-Smith.
• For unmatched forms, require exact surface-form lemma agreement between MorphGNT and PROIEL, or a registered contextual/source-native rule with corroborating morphology.
• Use pinned Tischendorf morphology only as an additional annotation witness; record its shared textual ancestry with PROIEL.
• Withhold English whenever word alignment or lexical identification is ambiguous.
• Exclude OCR and AI image transcription from certification.

Not permitted

• NA28 text presented as Vaticanus
• Silent resolution of corrections or uncertain letters
• Filling a Vaticanus omission from another tradition
• Crossing English meanings from another tradition column
• Using OCR or AI image transcription to certify English
• Treating Parker/Heinfetter as a diplomatic representation of GA 03

Next certification action: Retain the three remaining non-lexical or incomplete manuscript forms as explicit red manuscript-status annotations unless stronger Vaticanus-specific evidence emerges.

Governing rules

• Display the pinned GA 01 source, never a critical-edition proxy.
• Select the base reading: original scribe including recorded first-scribe corrections; exclude later a-c correctors from display.
• Preserve damage, supplied text, abbreviations, correction layers, textual omissions, and comparison gaps as distinct states.
• Use Anderson 1918 verbatim; the system assigns published words but never translates.
• Assign zero to three consecutive English words to each extant Greek parent with deterministic monotonic ordering and fixed tie-breaking.
• Suppress Anderson wording with no extant GA 01 parent rather than masking a manuscript omission.
• Keep every decision inside its canonical verse.

Not permitted

• Westcott-Hort or another critical text presented as Sinaiticus
• Combining later corrector hands without attribution
• Generated or AI-translated English
• English-only display rows that conceal a GA 01 omission
• Cross-verse spans
• Merged or continuation cells
• Arrow glyphs

Next certification action: Retain the pinned source and rerun the source and Anderson certificates whenever Sinaiticus cells, source files, or alignment rules change.

Governing rules

• Retain source-token order.
• Give additional Latin words their own alignment rows.
• Use empty cells only where Latin has no corresponding word.
• Treat Whitaker as a lexical aid requiring contextual review.
• Keep published English in source-supported translation units; Urevangelium aligns but does not translate.

Not permitted

• Calling this text Weber–Gryson/Stuttgart
• Dating the displayed edition to 383 CE
• Dropping excess source words
• Presenting Whitaker lexical output as contextual translation
• Subdividing published English more finely than its source supports
• Silently harmonizing Challoner and Rheims

Next certification action: Obtain independent scholarly review of the Vulgate source audit, whole-unit English alignment method, and edition-specific adjudication ledger.

Governing rules

• Keep Greek and Latin sides distinct within the same codex witness.
• Require each displayed form to consume one unique occurrence in the corresponding TEI verse.
• Treat a shared row as comparative placement, not literal bilingual equivalence or diplomatic sequence.
• Represent full and side-specific physical loss separately from alignment and unpopulated display gaps.
• Within an apparatus element, use the first encoded rdg as the displayed base-reading stream while retaining alternatives in the TEI.

Not permitted

• Filling a physical lacuna from another tradition
• Treating blank post-generation rows as manuscript omissions
• Claiming comparison-row order is diplomatic sequence
• Flattening unused apparatus readings into the certified display

Next certification action: Retain the pinned TEI files and rerun certify:bezae:display whenever Bezae cells or source files change.

Governing rules

• Preserve every Syriac source token exactly once, with source-file hash, source occurrence, and stable row identity.
• Preserve RTL display while allowing a governed shared-row move to expose meaningful cross-tradition correspondence; the move never changes the stored Syriac token sequence.
• Keep every governing Murdock verse independent and preserve every admitted word verbatim.
• Assign zero to three contiguous Murdock words to each Syriac parent using fixed evidence priorities and stable tie-breaking.
• Use SEDRA IV only for atomic lexical headwords and ETCBC/SyrNT for morphology. Cross-tradition placement evidence requires two independent witness families among Greek, Latin, and Coptic; dependent Greek columns count as one family. Murdock remains the sole displayed English authority.
• Use an explicit English-only expansion only when all honest adjacent parents are at the three-word ceiling; leave a Syriac parent blank only when the published Murdock unit is mathematically shorter than the source-token inventory.
• Rerun source, English, row-parent, and governed-alignment certificates after any corpus change.

Not permitted

• Claiming BFBS 1905 without an exemplar chain
• Treating statistical support or phrase context as dictionary equivalence
• Using proportional English distribution or AI-generated translation
• Borrowing displayed wording from another tradition
• Normalizing samekh forms in the archival display
• Dropping or duplicating source or English words
• Merged cells, continuation cells, arrows, or cross-verse English spans

Next certification action: Seek independent Syriacist review of the 2,726 governed shared-row decisions and the 28 mathematically explicit edge cases while preserving the pinned source and deterministic certificates.

Governing rules

• Use the pinned Byzantine CSV as the sole text authority.
• Admit contextual TAGNT English only for an explicitly Byzantine-aligned token with RP2018 identity evidence.
• Display project-adjudicated TBESG or MorphGNT-lexicon output in orange.
• Require exact RP2018 surface and morphology identity before any English admission.
• Treat the column as an edited textform rather than a physical witness.
• Represent a verse absent from RP2018 as omitted rather than filling it from another Greek tradition.

Not permitted

• Silent generic TAGNT fallback
• Calling the edited textform one physical manuscript
• Applying physical-manuscript lacuna rules
• Borrowing readings or English from another Greek column
• Displaying unresolved generated English as source translation

Next certification action: Seek independent specialist review of the published ledger and adjudication rules; preserve the current pinned source versions until a separately audited migration.