Background
The bridge speaks three edge-feed-spec sibling protocols, but only Top-of-Book (validated) and Market-by-Order (added in #2) are wired up. Midpoint (frame magic 0x4D44) is not:
src/ingest/codec_midpoint.rs offsets are unvalidated draft (per CLAUDE.md), pinned only by self-consistent round-trip tests.
- There is no
FEEDS row (no confirmed live multicast endpoint), and no golden fixtures.
Scope
- Validate
codec_midpoint field layout against a live Midpoint frame hexdump (and the spec / publisher if available).
- Add a
FEEDS row once the live group/ports are confirmed (see edge-multicast-ref/docs/hyperliquid.md).
- Vendor a Midpoint golden fixture and add an E2E test asserting
midpoint messages flow with correct fields (mid, method, quality_flags, timestamps), reusing the existing harness.
Acceptance
Midpoint decodes are validated, a FEEDS row exists for a real endpoint, and an E2E midpoint test is green in CI.
Follow-up to #2.
Background
The bridge speaks three edge-feed-spec sibling protocols, but only Top-of-Book (validated) and Market-by-Order (added in #2) are wired up. Midpoint (frame magic
0x4D44) is not:src/ingest/codec_midpoint.rsoffsets are unvalidated draft (per CLAUDE.md), pinned only by self-consistent round-trip tests.FEEDSrow (no confirmed live multicast endpoint), and no golden fixtures.Scope
codec_midpointfield layout against a live Midpoint frame hexdump (and the spec / publisher if available).FEEDSrow once the live group/ports are confirmed (seeedge-multicast-ref/docs/hyperliquid.md).midpointmessages flow with correct fields (mid, method, quality_flags, timestamps), reusing the existing harness.Acceptance
Midpoint decodes are validated, a
FEEDSrow exists for a real endpoint, and an E2E midpoint test is green in CI.Follow-up to #2.