Historical reference. This page describes an earlier checkpoint. See the current application release and current engine qualification.
Frozen backend baseline — 0.3 beta
This is the immutable published predecessor. The stable v1 successor is recorded separately in the release notes. Its verification retains this manifest byte for byte and checks the explicitly declared successor hashes; the historical release is not rewritten.
The backend baseline is words-0.3.0-beta.1. Its source and data hashes are recorded in the machine-readable manifest. The signed release tag v0.3.0-beta.1 fixes the complete product tree and archive contents.
The qualified implementation comes from signed commit 18250decd9a28e0a385f9afa330c6b205defb4bb. Release preparation changes only the backend version constant; matching, parsing, formatting, diagnostics and data remain byte-identical to the fully tested checkpoint. The manifest records both the qualification identities and release source hashes.
At this release, npm run verify checked the frozen source/data/adapter hashes and then ran the regression suite. The release qualified 36 groups; later frontend tests extend that suite without changing the frozen backend. A backend change requires a new reviewed baseline and appropriate comparison against this release; do not rewrite this immutable tag or its published archives. Frontend work may continue independently. In the 0.3.1 frontend release, the application/package version is 0.3.1-beta.1; the frozen engine continues to report 0.3.0-beta.1. The website manifest records both identities.
The measured counts and exceptions are in backend acceptance. The reference is WORDS 1.99.0, snapshot words-mk270-1f2f0fb, from commit 1f2f0fb0867a896d7b9284a03d615ed635d6f992. The portable fixtures include the full English observations, source-selected option sessions and native boundary states. The fixed-wordlist corpus and reproduction route are identified in the acceptance report. The original source archive and licenses remain bundled.
Classical-text corpus acquisition and processing are deferred. This baseline provides a fixed comparator for that later work; its beta label does not expand the measured compatibility claim.