Status: DRAFT
Purpose:
Define the formal semantics and trust model for the future GLYPH Corpus Commitment Layer.
Core correction:
The goal is not to invent verifiable pattern matching from scratch.
Prior art already exists.
The GLYPH goal is:
hash-only, portable, replay-compatible, audit-amortized verification for byte-exact retrieval over static corpora.
Current chain:
Retrieval → Evidence → Bundle → Portable Replay
Next chain:
Evidence → Bundle → Commitment → Audit → Case
Core principle:
Replay is not removed.
Replay is amortized.
Old model:
every verifier replays every claim.
New model:
auditors replay each corpus once.
verifiers validate proofs forever.
Claim Types:
- Presence Claim
This exact byte sequence occurs in this exact corpus at this exact offset.
- Absence Claim
This exact byte sequence does not occur in this exact corpus.
- Count Claim
This exact byte sequence occurs exactly N times in this exact corpus.
- Interval Claim
This query corresponds to suffix-array interval [l, r).
Trust Model:
Publisher: creates corpus, index, commitment.
Auditor: rebuilds deterministic index from corpus and verifies commitment correctness.
Verifier: checks proof against commitment without rerunning full GLYPH retrieval.
Threats:
- Malformed suffix array.
- Wrong corpus fingerprint.
- Incorrect absence proof.
- Incorrect count proof.
- Encoding mismatch.
- Mutable corpus state.
- Overclaiming legal or semantic meaning.
Non-goals:
No semantic search. No embeddings. No ANN. No ZK requirement for V1. No confidential corpus proof. No blockchain-first design. No patent-dependent accumulator design.
Preferred V1 Direction:
Hash-only Merkle corpus + Merkle suffix array.
Reason:
- compatible with GLYPH
- implementable by one founder
- no trusted setup
- no pairings
- no patent exposure
- independently verifiable
- portable across machines
Next Required Design Documents:
- GLYPH_COMMITMENT_CLAIMS_V1.md
- GLYPH_COMMITMENT_TRUST_MODEL_V1.md
- GLYPH_COMMITMENT_ARCHITECTURE_V1.md
- GLYPH_COMMITMENT_AUDIT_PROTOCOL_V1.md