Repository path: workshop/translations/kasagi/collation.md · rendered 2026-09-09

«Kaşağı» — the copy-text, and why the choice is not a formality

The finding, in one sentence: the two freely reachable Turkish e-texts of this story are not two witnesses to one text but two different modernisations of it, disagreeing at 103 places in 826 tokens, and they modernise in opposite directions at at least one site — so no reachable text is the 1919 text, and the translator is choosing a modernisation state whether or not it notices.

1. The copy-text adopted

tr.wikisource.org, page "Kaşağı", revid 172873, fetched 2026-08-11, stored whole and unedited as source-trwikisource.txt with the paragraphs and dialogue lines numbered [1]–[74] (887 words). The only changes made in numbering it: the wikisource indent colons were dropped, — was written as —, and the category link was removed. Nothing lexical was touched.

Ömer Seyfettin died in 1920; the story was first printed in 1919, in Ottoman script. No scan or transcription of the 1919 printing was reachable, and the story is therefore worked from a modern Latin-alphabet recension whose editor is not named. That is a weaker footing than ARM-dakghar had (page images for every page) and is stated as a limit rather than absorbed.

2. The second witness, and what it turned out to be

A Turkish Ministry of Education school PDF (yeihoo.meb.k12.tr, a collection of Ömer Seyfettin stories; consulted 2026-08-11, logged in wiki/base/consulted.md) was fetched as a collation witness. Its embedded fonts carry no ToUnicode map, so the text was decoded from glyph ids — Latin range by the standard ordering, the twenty Turkish-specific and punctuation glyphs by an empirical map read off the opening sentence and checked through the whole story.

It is not a witness to the same text. Tokenised and aligned against the copy-text: 826 tokens against 819, 701 matching, agreement 0.852, and 103 divergence runs. The overwhelming majority are one lexical stratum being replaced by another:

The school recension also adds text — ahırı süpürmek ("sweeping out the stable") in the list of chores at [1], which the copy-text does not have — and drops ihtimal at [22].

3. The site that stops this being a simple story of an old text and a simplified one

At [5] the copy-text reads uyumlu and the school recension reads ahenkli — and here it is the school text that carries the older, Arabic-derived word and the wikisource text that carries the twentieth-century replacement. The modernisation runs in both directions, so neither e-text can be treated as the conservative witness and the other as its simplification.

[1] shows the same thing structurally: the copy-text's Dadaruh eline kaşağıyı alıp sits in a sentence the school text has rebuilt around an added chore.

4. What this binds

1. No page of this project may cite either e-text as "the 1919 text". The copy-text is a modern recension of unknown editorship and is named as such wherever it is cited.
2. Any locus this project manipulates or quotes is checked in both recensions first, and a locus where the two disagree lexically is not used for a measurement that turns on the wording. The study limb's thirteen loci were checked this way; E-20260811g-carrier §3 records the outcome.
3. The MEB recension is not stored. The divergence table above is the record; the full decoded text was held in scratch and not committed, because the simplification is an unnamed editor's work even where the underlying story is public domain (charter §7 hygiene).