CLAIM_BOUNDARY.md

Claim Boundary

HVCE v4 is intentionally aggressive but not dishonest.

What is claimed

1. HVCE v4 is a complete, runnable, portable Python reference implementation.
2. It round-trips files/directories exactly and verifies decoded bytes.
3. It adds real archiver features missing in earlier HVCE versions: password protection, recovery records, metadata preservation, GUI wrapper, Windows shell scripts, ZIP export, safe extraction, and benchmark tooling.
4. It can beat ordinary ZIP/Deflate by very large margins on small structured folders because it uses solid micro-packs and stronger modeling.
5. It can beat xz/zstd/ZIP on exact generated structured fields where its recipes match the data-generating law.
6. It can compress versioned high-entropy data by storing one raw state plus exact sparse patches when near-duplicate chunks exist.
7. It handles random/encrypted/already-compressed data correctly by avoiding futile modeling and falling back to raw/dedupe/patch behavior.

What is not claimed

1. It does not beat Shannon information theory.
2. It does not losslessly shrink arbitrary random data.
3. It does not decrypt encrypted files without the key.
4. It does not guarantee smaller output than every modern compressor on every corpus.
5. It does not claim security audit equivalence to mature cryptographic libraries.
6. It does not yet match mature desktop tools in installer polish, localization, signed binaries, years of fuzzing, or enterprise support.

Correct interpretation of “dominate random/encrypted/already-compressed areas”

For these areas, ratio dominance is impossible in the universal lossless sense. HVCE v4 dominates by archive behavior:

• fast recognition;
• low expansion;
• exact duplicate detection;
• sparse version patching;
• authenticated optional encryption;
• optional recovery;
• verified extraction;
• safe metadata-preserving storage.

That is the strongest mathematically defensible target.