Codex: cod. Mus. Brit. Add. 14,644 - Nestle's sub-head is 2) e codice Mus. Brit. Add.
14,644., printed mid-page on p. 113. Sub-piece IV.2, the second of the three named codices in
Chrest. IV; 274 lines over pp. 113b-126, spanning foll. 18r-23v.
⭐ The collation style CHANGES with the codex: IV.1 cites Paris. 234 by folio and column
((b, col. 2.)); IV.2 cites Add. 14,644 by folio and side only - (18 r.), (f. 18 vers.),
(fol. 19 rect.), (fol. 19 vers.) … (f. 23 vers.) - because the second manuscript is not in
columns. A marker regex written against IV.1 will match nothing in IV.2.
⭐ RULED - and this is the trap the record exists to close. Nestle's published corrigenda
in De sancta cruce ("Korrekturen zu meiner Abschrift von B: Z. 148 lies …, 216 …, 261 …, 343 …; zu
Z. 44 siehe die Praef. der Porta … z. B. Z. 222") do NOT touch Chrestomathia IV. In that
footnote A and B are MANUSCRIPT SIGLA, not De sancta cruce's section letters: A = Bodl. Libr. MS.
Marsh 13 (dated fol. 243v, A.D. 1177) and B = Mus. Brit. Add. 12,174 anni 1196 - the manuscript
behind DSC's first section, A.) (E COD. LONDIN. ADD. 12174 ANNI 1196.). That text runs to line
345/346, which is why both the corrigenda and the collation apparatus headed A [= Bodl. Libr. Ms.
Marsh 13] reach line 343 - a line number the Historia inventionis never reaches (IV.1 = 101 lines,
IV.2 = 274, IV.3 = 92).
⛔ The near-miss is the whole danger: 44, 148, 216, 222 and 261 are ALL valid line numbers inside
Chrest. IV.2, and 216 and 261 are exactly the lines that carry (fol. 22 vers.) and (f. 23 vers.).
A later session that "applies Nestle's own corrigenda" to IV.2 would silently corrupt five lines of
a text they were never written about. What makes it look as though they belong is that the
apparatus is printed physically AFTER our text - DSC pp. 37-38, "steht S. 37. 38 hinter dem Text
von B die vollständige Collation von A". It does not belong to it.
Declared from this plate, not from Phase 0's note and not from the neighbouring page. The whole of
Chrestomathia IV - all 24 pages, all three codices - sits at the unvocalized-Serto pole and never
leaves it: not one vowel point on any page. What the plates do carry is seyame on plurals, the
letter points (dalath/rish, and the qushshaya/rukkakha and participle points), a linea occultans
over elided letters, and the ܂ / ܆ / ❖ punctuation.
⭐ So Chrest. IV settles the layer table's third case. Chrest. I alternates by caput (Serto /
Estrangela / Serto / Estrangela) and puts two layers on three pages; Chrest. II is unvocalized
Estrangela throughout; Chrest. IV is unvocalized SERTO throughout - the same vocalisation state
as Chrest. II but the other script. Script and pointing are two independent axes and this piece is
the proof; a layer field that treats them as one string will keep colliding.
E. Nestle, DE SANCTA CRUCE: ein Beitrag zur christlichen Legendengeschichte, Berlin: Reuther 1889 - archive.org desanctacruceein0000nest (148 leaves; leaf = printed page + 11, calibrated on the plate at n33 = p. 22). Same passage, lifted by Nestle out of his own book of the same year: same section head B. HISTORIA INVENTIONIS SANCTAE CRUCIS, same sub-heads, same unvocalized Serto, same marginal numbering, same folio/column markers. Character-level Syriac diff target - but NOT an independent witness: agreement proves the keying of NESTLE, not of the manuscript. See research/syriac-peshitta-editions.md
seyame and the letter points are visible at the plate's own resolution and are reported; individual qushshaya/rukkakha points are NOT itemised (below the resolution at which this pass can be honest about them - a known gap, not a claim the plate lacks them).
Marginal, every fifth line, in the OUTER margin (left on versos, right on rectos) - verified on this plate. ⭐ The numbering is per sub-piece and restarts at each codex: IV.1 runs 1-101, IV.2 restarts at 1 and runs to 274, IV.3 restarts at 1 and runs to 92. A reader who takes the marginals as one series across Chrest. IV will mis-locate every line from p. 113 on. This page: 35, 40, 45, 50 - RIGHT margin (recto, outer). Line range 35-54. ⚑ Page line-ranges here are derived from the PRINTED marginals, not from a row count: on each page a marginal numeral pins one row to an absolute line number, and the next page's first marginal pins the page's last line. Every boundary in pp. 108-131 closes without a gap or an overlap.
⛔ DELIBERATE NON-TRANSCRIPTION, not a failed plate fetch. Reasons, in order:
1. Unvocalized Serto carrying an unknown text. R3(a) worked because it keyed texts it already
knew (Genesis, Matt. 5) and then checked them, and because Chrest. I is VOCALIZED Serto, where
the points disambiguate the skeleton. Chrest. IV is unvocalized Serto carrying a legend for which
no digital edition exists anywhere (ruled 2026-09-01, research/syriac-peshitta-editions.md).
Nothing primes the read; every word is a cold glyph-by-glyph decision.
2. The master scan is 1586 x 2512 px. Checked 2026-09-01: page/nNNN.jpg, the same URL with
?scale=0/1/2, and the _jp2.zip member all return that identical size, so there is no
higher-resolution copy of this scan to escalate to. Serto pairs that differ by one hook or one
dot (dalath/rish, the yud/shin/semkath peaks, medial nun vs. lamadh vs. alaph) sit at the edge of
it; past ~2.5x the enlargement is interpolation, not information.
3. The house rule forbids the alternative. [[reference_arabic-control-rule]] and
SYRIAC-LANGUAGE-PILOT.md §5/§8: never read codepoints by eye; an extractor that is unsure flags
rather than extracting "to be safe". A confidently wrong Syriac line is worse than a blank one.
4. Keying from the diff target would poison the diff. De sancta cruce is the clearer of the two
settings (1726 px scan, larger type on the page). Keying from IT and then "diffing" against the
Grammatik would make the agreement an artefact of the method - the circularity R3(a) refused when
it declined to peek at BFBS for Matt. 5:5.
⭐ What is owed here is a reader with Syriac, or a Serto-trained HTR model, working from De sancta cruce and the Grammatik side by side. Everything else in this record - line numbering, the folio/column collation, the codex boundaries, the footnote apparatus, the De sancta cruce line-for-line correspondence, the corrigenda ruling - is done and does not need redoing.
⛔ NOT READ - the Syriac of this page is deliberately not transcribed (see transcription_note).
The line inventory below is what the plate does license: one entry per printed line, with the line's
number in the piece's own marginal series and anything on that line that is not Syriac.
Line inventory:
line 35: marginal numeral 35 printed
line 40: marginal numeral 40 printed
line 43: (fol. 19 rect.)
line 45: marginal numeral 45 printed
line 50: marginal numeral 50 printed
NONE. Nestle prints no footnote on this page. ⭐ And that holds for the whole of IV.1 and IV.2: not one philological footnote in the nineteen pages 108-126. The only apparatus in those two sub-pieces is the marginal line numbers and the folio/column markers. Nestle's notes on Chrest. IV appear only in IV.3 (pp. 128, 129, 130), which also carries in-text editorial parentheses. ⚑ This narrows a finding from R3(a): the Chrest. II record's remark that Chrest. II has 'no philological footnotes at all, unlike Chrest. IV' is true only of Chrest. IV.3 - 19 of Chrest. IV's 24 pages have no footnotes either.
⛔ Syriac not keyed (see transcription_note). What IS aligned - exactly, and verified on independent anchors - is this page's LINE NUMBERING and its folio/column markers against Nestle's own De sancta cruce (1889), the same setting.
⭐ The two printings share their line breaks and their line numbering - verified, not assumed.
De sancta cruce sets more lines per page than the Grammatik (DSC p. 21 carries text lines 1-20 where
Grammatik p. 108 carries 1-17), so the PAGES do not correspond - but the LINES do, one for one, and
a given line carries the same marginal numeral in both books. Anchors, in both sub-pieces:
· IV.1 (a, col. 2.) = line 12 in both · (b, col. 1) = line 22 in both
· IV.1 (b, col. 2.) = line 80 in both (the DSC line carries the marginal 80 itself)
· IV.1 (fol. 295, a, col. 1.) = line 92 in both
· IV.2 (f. 18 vers.) = line 18 in both · (f. 23 vers.) = line 261 in both
· Both texts end on the same line: IV.1 at 101, IV.2 at 274 with the same closing mark.
That correspondence is what makes a character-level diff possible at all, and it is the half of the
diff this pass can honestly deliver. It also hands the next session a free locator: any line N of
Chrest. IV.1/IV.2 is found in De sancta cruce by its own marginal number, with no page arithmetic.
IV.1 = DSC pp. 21-25, IV.2 = DSC pp. 26-36.
⛔ The character-level diff itself is NOT RUN. With the Grammatik side unkeyed there is nothing to compare glyph for glyph, and running it against DSC alone would only re-key DSC. Recorded as owed, not as done. ⚑ When it is run, an agreement will prove only that both printings were set from the same transcript - never that the transcript matches the manuscript ([[feedback_blind-reader-fact-vs-rule]]: transcription check, not textual check).
⭐ On this page specifically: line 43 carries (fol. 19 rect.). No structural divergence was found anywhere in
pp. 108-126 - not one line-number offset, not one displaced or missing folio marker, in 19 pages.
That is a real result and it is the strongest thing the same-setting target can give short of the
character diff: it means the Grammatik's Chrest. IV is a straight re-impression of De sancta cruce's
section B, not a re-set text, so a later character diff should expect compositor-level divergence
only and can treat any word-level divergence as a signal.
⭐ RULED - and this is the trap the record exists to close. Nestle's published corrigenda
in De sancta cruce ("Korrekturen zu meiner Abschrift von B: Z. 148 lies …, 216 …, 261 …, 343 …; zu
Z. 44 siehe die Praef. der Porta … z. B. Z. 222") do NOT touch Chrestomathia IV. In that
footnote A and B are MANUSCRIPT SIGLA, not De sancta cruce's section letters: A = Bodl. Libr. MS.
Marsh 13 (dated fol. 243v, A.D. 1177) and B = Mus. Brit. Add. 12,174 anni 1196 - the manuscript
behind DSC's first section, A.) (E COD. LONDIN. ADD. 12174 ANNI 1196.). That text runs to line
345/346, which is why both the corrigenda and the collation apparatus headed A [= Bodl. Libr. Ms.
Marsh 13] reach line 343 - a line number the Historia inventionis never reaches (IV.1 = 101 lines,
IV.2 = 274, IV.3 = 92).
⛔ The near-miss is the whole danger: 44, 148, 216, 222 and 261 are ALL valid line numbers inside
Chrest. IV.2, and 216 and 261 are exactly the lines that carry (fol. 22 vers.) and (f. 23 vers.).
A later session that "applies Nestle's own corrigenda" to IV.2 would silently corrupt five lines of
a text they were never written about. What makes it look as though they belong is that the
apparatus is printed physically AFTER our text - DSC pp. 37-38, "steht S. 37. 38 hinter dem Text
von B die vollständige Collation von A". It does not belong to it.