nx_meal_source.nx

nx_meal_source.nx -- LIB: the LINK-ROT-PROOF provenance spine of the meal planner. Every recipe rendered on a Nishi page carries WHERE IT CAME FROM -- publisher, original URL, license -- together with a PRESERVED COPY of that origin page written as a WARC/1.0 resource record (nx_web_archive, ISO 28500 -- the same format the Internet Archive and Browsertrix emit, so ours interoperates). When the origin rots, the citation still resolves: the page serves the preserved bytes instead of a 404, and the credit line stays TRUE instead of pointing at a dead host. HONESTY BY CONSTRUCTION (rule 23; R0a "a placeholder that reads as a real claim"): ms_is_archived is a MEASUREMENT, never a stored flag -- it re-reads the WARC and confirms the payload is actually retrievable before the page is permitted to render "archived copy". A recorded INTENT to archive can therefore never masquerade as an archive that EXISTS. ms_capture_verify proves the same for a fresh capture, round-trip, byte-for-byte -- binary payloads with embedded CRLF are exactly where a line-based archive corrupts silently. Schema (prefix passed in, e.g. knowledge/store/meal-): meal:srcids -> TAB list of recipe ids meal:src:<rid> -> title <t> publisher <t> origin_url <t> license <t> captured_day <t> origin_bytes This file RECORDS and MEASURES provenance; it renders NOTHING (rule 9) -- nx_meal_page owns all markup. license_tier: ORIGINAL No hw writes (Rule 26).