Skip to content

Semantic Fingerprints and Integrity

v9 note: V8 semantic fingerprints remain valid under their profiles. They are not retroactively replaced by v9 Logical Commitments; see Identity-and-Commitments-v9.

v8 separates byte/content identity from semantic fingerprinting.

SEMANTIC FINGERPRINT != BYTE HASH

Byte identity

Inherited v6/v7 content hashes continue to identify canonical bytes under their declared canonicalization boundaries. v8 does not change those rules.

Semantic fingerprint

A semantic fingerprint is an optional derived digest over a versioned semantic projection. It is useful for caching, deduplication and query continuation when irrelevant presentation differences should not change the semantic key.

A fingerprint must declare a profile, for example:

kristal.semantic-fingerprint/v1

The profile defines included semantic fields, excluded presentation/derived fields, canonical ordering, scalar encoding, hash algorithm and domain separator.

Processors cannot invent local exclusion lists while claiming the same profile.

Language exclusion

Canonical semantic fingerprint profiles should normally exclude external lexical labels when those labels do not alter meaning. Lexical Kristals have their own content/fingerprint identities.

Tamper evidence

A local hash chain or Merkle root is not independent proof if the same attacker can rewrite both data and chain head. High-assurance deployments should anchor release/snapshot roots in an independently governed system and define restore/fork reconciliation.

Authority

A valid digest proves identity/integrity under its profile. It does not prove factual correctness or epistemic authority.