Corpus gate
Status: open/active · Aliases: byte-identity gate
Definition
Section titled “Definition”An empirical CI release gate for fidelity claims: a large body of real files that every build must round-trip byte-identically — parse, serialize with no edits, byte-compare; any single-byte diff fails. No-op-edit variants (apply an edit and its inverse, require zero diff) extend it to writers. Losslessness is unprovable in general — edge cases are unbounded — but it is testable against the files you own, so the gate, not spec-completeness, defines done. Precedent: rumdl’s 660-file byte-identical corpus from D-0011-markdown-formatting-tool.