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

「祝盃」 — copy-text collation against the 1921 first book printing

Run late, and the lateness is the first thing to say. T-shukuhai-R06-v1 was frozen at 48f2856 declaring one witness and saying that no second freely reachable text of this story was found. That statement was false and is corrected here. The second witness was located after the translation and the whole of E-20260812d had been frozen and run — so this collation could not have changed any design, and it did not. What it can do is say whether anything measured sat on a disputed reading.

The two witnesses

The 1920 first printing (『電氣と文藝』, September 1920) is not digitised anywhere reachable. NDL Search holds two catalogue records for the magazine, both paper-only, with holdings at the Kanagawa Museum of Modern Literature. Consulting it would require a physical visit. So this is a 1960 text against a 1921 text, not against the first printing.

Method

All six page images fetched at full resolution and read directly, right to left, against the 45 numbered units of the copy-text. Orthographic modernisation was expected throughout and is not recorded: 居た→いた, 云ふ→いう, 默→黙, 學→学, 臺灣→台湾, 這入つた→入った, 丈→だけ, 就いて→ついて, 遉に→さすがに, 喰→食, ／＼ iteration marks, and historical kana generally. Only differences that survive normalisation are listed.

What the collation found — five divergences, none on a measured site

[11] is the one that matters, and it lands on a sentence the translator had already stopped at. T-shukuhai-R06-v1's log records at D12:

"「[11] 何の意味もなかった」 — 'neither of them meant anything by it' — the Japanese is literally 'there was no meaning in it for either of them' … Both readings are live; I took the second because the paragraph goes on to give the feeling the words did not carry."

The ambiguity the log spent an entry adjudicating does not exist in the 1921 text, which reads 興味 — interest — and settles it a third way that was not among the two options considered. The translator was choosing between two readings of a word the author may not have written.

The English as frozen is "But neither of them meant anything by it"; on the 1921 reading it would be "But neither of them had any interest in it." The translation is not corrected: it is a frozen artifact of a stated copy-text, and altering it after the fact would destroy the record. The divergence is recorded here and on RS-20260812d §10.

Bearing on E-20260812d

None of the five divergences falls on a manipulated site, and every one of the 26 items' edited strings was checked individually against the 1921 page images and found present:

• All 15 selected windows' sites are attested in 1921, verbatim under normalisation — [22] 止められなかった, [3] 埋もれてしまった, [24] が、久野も啓吉も、それ以来の, [4] 啓吉も久野も 文壇的には, [12] 啓吉も久野も話し出さなかった, [25] 啓吉の心を暗くした, [20] 再校は向うで 見て呉れるよ, [31] もう会わないかも知れないよ, [38] ちっとも変っていないじゃないか, [43] 俺達は考えて見ると, [14] 僕の本ね, [18] 僕のも同じだった, [28] 君は、この電車に, [34] おい！　ソーダ水でも, [40] でも、分らないものだね.
• All six control items' sites are attested — [1] 五六年前, [36] 啓吉も、久野の気持が 分った, [45] それは、ソーダ水であった, and the two byte-identical nulls.
• Three of the five divergences — [6], [7], [11] — sit inside windows the Stage-S selection had already dropped (BS2 and BN4). [24] and [39] sit in selected windows (FS3, FV4) but in the surrounding context, not at the edited string.

So the run's measured material is not in dispute between the two witnesses. That is a fact established after the fact and by luck, not by design, and the honest statement is that this project translated and measured a text for a full session before checking whether it was the text.

What a future session should not repeat

The reachability search that found this witness was run after the freeze, to measure contamination. It should have been run before, as the copy-text gate, exactly as it was on 「ডাকঘর」 (S160, S165) and 「Kaşağı」 (S161). The Aozora card names its 底本 — a 1960 collected edition — on its face, and a 1960 collected edition is a signal to go looking, not a copy-text to accept. Method note (bmn).