# KIP-0024: Byte-Path Fixture Freeze Milestone 18 stabilizes deterministic byte-path fixtures before the project adds protocol-feature corpus work or broader wire-shape generation. Milestone 19 builds on this fixture layer with [KIP-0025](KIP-0025-protocol-feature-corpus.md). The goal is to make future byte-path drift visible without storing raw payloads, raw encoded bytes, ciphertext, secrets, keys, nonce bases, auth tags, proof material, or destination data. ## Purpose Milestone 17 proved that local adapter/runtime/carrier output can move through a deterministic byte encoder, bounded byte pipe, decoder, sequence checker, and receiver-side reconstruction model. Milestone 18 freezes safe summaries of that path so future changes can be compared against known baselines. ```text profile seed -> compiled profile -> local adapter scenario -> runtime session -> proxysem + carrier + security metadata -> byte transport encoder -> deterministic byte pipe -> byte transport decoder -> receiver reconstruction -> safe fixture summary -> generated/interpreted parity check ``` ## Fixture Manifest `internal/fixtures` defines a versioned manifest with: - fixture schema version - deterministic generation timestamp - fixture set name - profile seed list - scenario list - backend version - payload/secret hygiene flags - fixture entries - safe byte-path summaries - optional malformed corpus metadata - optional performance baseline buckets The committed golden set lives in: ```text testdata/fixtures/bytepath-golden.json ``` The seed set is: ```text 12345 12346 12347 ``` The scenario set is: ```text byte_single_flow_echo byte_many_small_flows byte_large_flow_fragmented byte_mixed_flows byte_reset_isolation byte_corruption_rejection byte_replay_rejection ``` ## Safe Hashes Fixture entries include: - `summary_hash` - `trace_hash` - `byte_shape_hash` These hashes are computed over safe, normalized metadata. They are not hashes of raw payloads or raw encoded byte streams. ## Malformed Byte Corpus `testdata/fixtures/malformed-byte-corpus.json` stores metadata for malformed-input regression cases, including expected reject buckets. The actual malformed byte sequences are constructed deterministically by tests and gates. Covered cases include: - empty input - truncated header - unknown frame kind - oversized frame - oversized payload - invalid length field - trailing invalid bytes - invalid fragment metadata - duplicate or missing fragments - oversized reassembly - sequence replay - old or future sequence - post-close/post-reset data - corrupted checksum - invalid or oversized metadata - random bounded noise ## Generated And Interpreted Parity `internal/byteparity` compares interpreted and generated byte-path summaries for: - profile ID - scenario name - frame count - fragment count - reassembly result - sequence rejection result - malformed rejection result - corruption rejection result - replay rejection result - backpressure bucket - runtime stream mapping count - target error/reset count - sink completion - payload/secret hygiene Semantic drift fails. Safe byte-shape differences are reported separately so generated specialization can be inspected without being confused with semantic failure. ## Fixture Drift Detection `kcheck fixtures verify` reloads the committed golden manifest, regenerates the deterministic fixture set, and compares hashes. Unexpected drift fails. `kcheck fixtures compare` compares two manifests and reports added, removed, changed, semantic-drift, and byte-shape-drift entries. ## Performance Baseline Buckets `testdata/fixtures/performance-baseline.json` stores broad warning-only buckets for: - byte frame encode/decode - byte pipe write/read - local adapter scenario - byte transport quick audit - fixture verification These are intentionally broad to avoid flaky exact timing thresholds. ## Trace Hygiene Rules Fixture and parity outputs must not contain: - raw payloads - raw encoded or decoded bytes - ciphertext or plaintext dumps - auth tags - nonce bases - secrets or derived keys - client/server write keys - proof material - private keys - session secrets The hardening trace hygiene scanner includes fixture manifests, parity reports, malformed byte corpus metadata, and generated fixture outputs. ## Audit Gates `kcheck bytepath` reports: - `fixture_bytepath_drift` - `bytepath_fixture_stability` - `bytepath_generated_interpreted_parity` - `bytepath_malformed_corpus` - `bytepath_regression_baselines` - `bytepath_fixture_trace_hygiene` - `bytepath_performance_baseline` The default quick audit also includes the required bytepath gates. ## Commands Generate the byte-path golden manifest: ```bash go run ./cmd/kcheck fixtures generate --out testdata/fixtures/bytepath-golden.json --force ``` Verify committed fixtures: ```bash go run ./cmd/kcheck fixtures verify ``` Compare manifests: ```bash go run ./cmd/kcheck fixtures compare --old testdata/fixtures/bytepath-golden.json --new testdata/fixtures/bytepath-golden.json ``` Run bytepath audit: ```bash go run ./cmd/kcheck bytepath --quick go run ./cmd/kcheck bytepath --full --out testdata/audit/bytepath.json ``` ## Generated Backend Parity Generated modules include: ```text protocol/bytepath_fixture_test.go protocol/bytepath_parity_test.go ``` The generated source scanner checks for bytepath fixture constants and tests. `kcheck codegen --quick` includes `bytepath_fixture_generated_backend_parity`. ## Limitations The fixtures freeze local deterministic byte-path metadata only. They do not prove real-world transport behavior, undetectability, or production readiness. They do not store or evaluate raw packet captures. Future wire-shape work needs a separate corpus and review. ## Next Milestone Milestone 19 adds a protocol-feature corpus and wire-shape evaluation baselines using these byte-path fixtures as stable regression anchors. The next step is a wire-shape generation prototype that consumes those baselines.