Tags: everanium/itb
Tags
v0.2.0 — Lock Batch mode. Parallel wrapper. Binding fleet refresh Major changes - Lock Batch — the performance variant of Lock Soup overlay; opt-in via SetLockBatch(1) on both encrypt and decrypt sides under the PRF assumption. Wire is not interoperable with ciphertexts produced in Lock Soup mode without Lock Batch mode or vice versa. - Format-deniability wrapper now multi-core across all entry points (one-shot and streaming, Go-native and every binding through libitb.so) for buffers at or above the 256 KiB threshold; smaller chunks stay on the single serial keystream. - Binding fleet — bench harnesses rewritten lazy (one case at a time) to bound peak memory; binding wrapper benchmarks adopt the canonical-order table layout mirroring the Go-side wrapper/BENCH.md. Compatibility note (ABI) - hashes.Registry primitive order is changed to the canonical binding order (Areion-SoEM-256/512, BLAKE2b-256/512, BLAKE2s, BLAKE3, AES-128-CTR, SipHash-2-4, ChaCha20). ITB_HashName(i) / ITB_HashWidth(i) callers iterating by index see the new mapping. Callers using the named string surface (HashName / Find) are unaffected.