Repository path: workshop/translations/atsui-suna/provenance.md · rendered 2026-09-09

熱い砂の上 — copy-text, provenance, and the contamination gate

The work

牧野信一 «熱い砂の上» (Makino Shin'ichi, On the Hot Sand), 1935. First published 「週刊朝日 第十六巻第二号」朝日新聞社 (Shūkan Asahi 16.2). Makino Shin'ichi, 1896–1936; the work is in the public domain in Japan and is served by Aozora Bunko.

A beach on a day of exceptional heat. Five bathers bolt across the scorching sand; two men showing off for a pair of geisha make it four or five yards and bolt too; a father runs with his child on his back; a troop of schoolboys start themselves with an imitated pistol. Everyone crosses. The narrator, left alone on the near side, turns the hot sand into an image of his own life, is called by his friends, runs, and is laughed at the whole way over. 3,198 non-whitespace characters, 106 paragraphs, four numbered sections.

Copy-text

• Retrieved: https://www.aozora.gr.jp/cards/000183/files/45281_39787.html, Shift_JIS, decoded as cp932, by tools/fetch_aozora.py on 2026-08-15.
• 底本: 「牧野信一全集第三巻」筑摩書房. 初出: 「週刊朝日 第十六巻第二号」朝日新聞社.
• 入力 宮元淳一 · 校正 門田裕志.
• Ruby readings, inline tags and Aozora annotation markers stripped by the fetcher; 0 gaiji notes, 0 gaiji images, so nothing was lost to substitution.
• Historical kana orthography (歴史的仮名遣) and the iteration mark ／＼ are preserved as printed. Both are load-bearing: ／＼ is the source's own reduplication marker and is a device class in device-sites.md.

Single witness, and it is declared rather than glossed over. The project's practice on a long work is to collate two free witnesses (ARM-legend's copy-text gate). No second freely reachable witness of this text exists — Makino is not in Project Gutenberg, has no Wikisource text, and the Chikuma Zenshū the Aozora file is set from is not online. The copy-text is therefore uncollated, and any conclusion that turned on a disputed reading would be unsupported. Nothing in the design turns on one: the manipulation is applied to the English, and both arms of every pair descend from the same single reading.

The contamination gate — run before any locus was selected and before anything was translated

CLAUDE.md's standing rule requires a tools/dependence_check.py measurement on the candidate material before an experiment is designed on it, and requires that where no comparator is reachable this is said on the artifact.

No comparator is reachable, and that is the finding. Two searches on 2026-08-15:

1. Makino Shin'ichi English translation short story — returns the author's biography and one English text: fragments of a different story, «ゼーロン» (1931), posted on a personal Medium blog. Nothing else.
2. «熱い砂の上» / "On the Hot Sand" translation English — returns an audiobook reading of the Japanese and no English rendering of any kind.

Makino is one of the substantially untranslated Japanese modernists. No published English rendering of this story exists that the project can find or reach, so the overlap statistic that dependence_check.py computes has no second operand and was not run. This is the case the standing rule anticipates: "where no comparator is reachable, say so on the artifact."

What this does and does not buy. It buys the absence of the specific mechanism RS-20260814f §8 and note (bhb) measure — a rendering walking back into a published hand it has read or remembers. It does not establish that the lead has never seen this text; it establishes that if it has, nothing exists against which the resemblance could be measured. The declaration on every artifact of this work is therefore contamination: none, with that basis, and the limit is stated on the result page rather than buried here.

The second contamination direction, which does have a measurement. The three arms are all lead output and the second and third are derived from the first, so their mutual overlap is total by construction and means nothing. Note (bhb)'s worry — the lead matching itself across independently written renderings — does not apply, because nothing here was independently written. The pair is a manipulation of one text, and the design says so.

Copyright hygiene

Public domain in Japan (author d. 1936); Aozora Bunko serves it under its own terms. The source is stored whole in source.txt because it is public domain, per charter §7. No copyrighted material was consulted; there is nothing to log in wiki/base/consulted.md.