nx_base16_extvec_gate.nx

nx_base16_extvec_gate.nx -- validated against RFC4648-10-BASE16, read from the pinned+corroborated RFC4648 document. RFC 4648 section 10 publishes SEVEN BASE16 vectors; section 8 specifies the Base16 alphabet as "0123456789ABCDEF" -- UPPERCASE, normatively. runtime/nx_hex_codec.nx's hex_encode emits LOWERCASE. ★★★★★ THE TEMPTING MOVE WAS THE GAMING MOVE, AGAIN: lowercase the authority's expected string before comparing, and this gate goes GREEN in one line. That is WEAKENING THE AUTHORITY TO FIT THE IMPLEMENTATION -- the same substitution as writing an Argon2 gate at p=1 because our Argon2 only does p=1. The published answer does not get edited to match us. SO THE GATE LOCALISES INSTEAD OF MERELY FAILING. Two tests per vector: T-EXACT byte-for-byte vs the document -> answers "are we RFC 4648 Base16 conformant?" T-NOCASE ASCII-case-insensitive comparison -> answers "is the NIBBLE MAPPING right, and only the alphabet case wrong?" ★A FAILING TEST THAT CANNOT SAY *WHICH PART* FAILED SENDS SOMEONE TO REWRITE THE WHOLE ENCODER. Splitting the comparison turns "our hex is broken" into "our hex is correct but is not Base16", which is a one-line remedy (a second, uppercase-alphabet entry point) rather than a rewrite. ⚠SCOPE: this gate binds `nx_hex_codec.nx`. A SECOND `hex_encode` with the same name exists in `hex.nx` -- the THIRD duplicate-library pair found in this subtree today (after sha1/nx_sha1 and hkdf_sha1/nx_hkdf_sha1). ★NAME THE FILE YOU BOUND, NOT THE FUNCTION YOU CALLED. Document pinned to a CORROBORATED digest: the sovereign fetcher and .NET WebClient independently agree (knowledge/extvec/CORROBORATION.tsv), so the expected values are the IETF's bytes, not ours. license_tier: ORIGINAL expect_exit: 0