D023 - The suffix is committed and labelled

Status: partly-superseded

byte-order half superseded by D024 · 2026-08-19

data/hash-suffix.toml carries the value. nid.py defaults to it; --suffix-file overrides.

Reverses the first half of D022, which had it supplied at build time and never committed. The emulator side of this project reached the opposite conclusion with better reasoning, and it is worth restating rather than merely deferring to:

• It is not a key. It decrypts nothing, signs nothing, protects nothing. It salts a name-mangling hash so the mapping from names to identifiers is non-obvious.
• Every emulator of this target necessarily contains it, because resolving imports is the central act of high-level emulation. Withholding it protects nobody.
• A labelled constant with its derivation written beside it is better provenance than no constant at all. The thing that makes a hex blob awkward to account for is the absence of an explanation, not its presence in the tree.

The second half of D022 stands: nid.py selftest exercises the encoding against arbitrary values and needs no real suffix, so the transform stays testable independently of whether this particular value is right.

The byte-order claim here was wrong - see D024

This entry originally asserted big-endian on the strength of a prose comment. That is disproven by a published test vector. The reasoning is left in place because the way it went wrong is the instructive part.