Interlanguage and Mathematical Translation Methodology Sidecar

Interlanguage and mathematical translation methodology sidecar Open first: 00_Interlanguage_Methodology_Current_v14_20260805.pdf remains the default overview. This concept is the public methodology, provenance, decision-rationale, correction/reversal, corpus-control, and continuation shelf; mathematical readers remain on their dedicated concepts. Purpose: this concept is the methodology, provenance, corpus-control, decision-rationale, and programme-state sidecar for the mathematical transcription and translation archive. It preserves source-body baselines, workflow controls, correction and reversal histories, continuation state, reproducible checks, and bounded interlanguage experiments; mathematical readers remain on their author or series concepts. Current corpus map: EGA remains on concept 10.5281/zenodo.20414353; SGA remains on concept 10.5281/zenodo.20410947 (current version 21810316); FAC remains on concept 10.5281/zenodo.21720996 (current version 21783868); GAGA remains separate on concept 10.5281/zenodo.21781322; and the Noether multilingual corpus remains on concept 10.5281/zenodo.20412587 (current version 21785492). EGA I canonical-French / paired-English checkpoint: the direct p.142 surface preserves the coherent diplomatic French and paired English source through printed p.142, the exact R90 English source tree, the French-rooted pre-Stacks indexing scaffold, privacy-clean French and English human logbooks through p.141, the p.142 French/English/workflow decision-reversal-error ledgers, continuation, status, privacy, rights, manifest, and validation surfaces. The next cursor is printed p.143, continuation of Proposition 6.1.10. Commit-pinned checkpoint 05690f56df35a596b578c4d89c58a840b10947c1. Translation as correction and quality assurance: translation is treated as a second source-audit layer, not merely as output generation. Re-expression in another language exposes ill-typed arrows, unbound symbols, broken references, grammatical defects, and hidden ambiguities that can survive transcription and monolingual review. Every proposed repair remains separately identified and adjudicated against the scan and diplomatic source; the diplomatic edition is never silently rewritten, rejected proposals remain visible, and accepted readings propagate reversibly to corrected and translated editions. The current SGA 1 case study records 238 translation-assisted candidates, of which 236 were accepted and two rejected. The complete evidence is in SGA record 21810316. This measured workflow supports quality assessment but does not replace direct source inspection, mathematical review, or the preserved diplomatic text. Quality boundary: archive recovery independently replayed the source manifests and built only bounded English and French p.142 wrappers. It did not run a global reader build, render pages, run OCR, perform translation or transcription, or claim mathematical/source-fidelity certification. Exact private source custody remains unchanged; every minimal public privacy transformation is ledgered. FAC trust evidence: the FAC project logbook, editorial decision logbook, and 219-entry self-correction/reversal ledger remain direct. The coordinator did not know the Achinger-Krupa translation existed when Codex translated and froze FAC nos. 1-79; the comparator was discovered afterward and all 79 blind-scope units were adjudicated against the French authority. Nos. 80-81 are outside the blind claim; the result is qualitative evidence, not certification or a scalar ranking. Korean Noether and Spanish SGA 5: the complete cross-corpus provenance bundle preserves the bounded Korean producer snapshots and their source-custody, checker, structural, difficulty, error/reversal, and continuation surfaces with honest UNCHECKED/uncompiled/unrendered/unassembled/unreviewed labels. Spanish SGA 5 is public on the existing SGA concept with its authorization, source-and-rights, continuation, and supersession records. These state labels are metadata, not release holds. Complete provenance bundle: 09_ARCHIVE_PROVENANCE__00_FAC_KOREAN_SGA5_COMPLETE_20260804.zip retains 478 outer members plus 17 recursively checked nested ZIPs / 622 nested members, including adverse and superseded histories. Private user-root path hits are zero; public attribution names are retained. Archive organization: every human logbook, decision/reversal/error history, continuation record, rights note, reader, and new EGA p.142 surface remains direct. Twelve older transport ZIPs are retained byte-for-byte inside 10z_Retained_Transport_Containers_20260804.zip solely to respect Zenodo's 100-file ceiling; its embedded manifest binds source URLs, bytes, MD5, and SHA-256, and predecessor record 21785605 retains their direct presentation. No distinct content is selected away. Limits: these model-built methodology, corpus, translation, provenance, and visual-QA artifacts do not claim native-language validation, source-fidelity certification, mathematical correctness, rights clearance, critical-edition status, peer review, or completion of any author corpus. CJK mathematical notation/backend: this version adds component 1.0.0, an evidence-backed shared document-mechanics layer with independent Simplified-Chinese, Japanese, and South-Korean locale profiles. It standardizes formula/source-note serialization, typed provenance, font and ToUnicode handling, deterministic builds, dual extraction, and render validation; it does not construct a merged CJK language or allow one language to authorize another. The readable component is commit-pinned on GitHub. The predecessor transport ZIP is preserved byte-for-byte inside 10z_Transport_v15.zip, and the other 99 top-level files are unchanged.