nx_otk_extvec_gate.nx

nx_otk_extvec_gate.nx -- NINTH provably third-party-validated claim: Poly1305 one-time-key derivation (not just the block function) vs RFC 8439 2.4.2, from the same pinned document. ★THIS IS THE SECTION THAT BROKE THE PREVIOUS PARSER, AND THAT IS WHY IT IS WORTH DOING. 2.4.2's ASCII gutter is ENGLISH -- "...would be it." -- and `be` is a whitespace-delimited, exactly-two-character, ALL-HEX token. The content-classifying reader I first shipped would absorb it as byte 0xbe and shift every byte after it. The LINE-BOUNDED reader used here takes token 0 as the offset, tokens 1..16 as data, and ignores the rest BY POSITION, so the gutter cannot contribute no matter what it spells. ★★★★LAW: A DISCRIMINATOR PROVEN ON ONE INSTANCE OF A DIALECT IS NOT PROVEN ON THE DIALECT. ★STRONGER THAN THE POLY1305 GATE: there, the message was ASCII prose so I supplied the input and let the published tag verify it. Here BOTH the plaintext AND the ciphertext are hexdumps in the document, so every byte on both sides of the comparison is read from the authority. Nothing is agent-supplied. license_tier: ORIGINAL expect_exit: 0