Conformance archive

Every Gateway API conformance run aether has written up, newest first — 10 of them, from the first diagnostic run that could not reach the traffic phase at all to the fully conformant one. Each is published as it was written, at the score it recorded that day. The verdict column is the run's own headline, not a re-reading of it.

The newest run is also served at /conformance/latest/; what CI enforces today is on the supported features page.

Date	Revision	Run	Verdict
2026-06-27	rev21	rev21: GATEWAY-HTTP FULLY CONFORMANT 43/43	GATEWAY-HTTP fully conformant; MESH-HTTP unchanged (3 known fails)
2026-06-26	rev20	rev20: GATEWAY-HTTP-Core 33/33 COMPLETE	Core 33/33 COMPLETE, the two expected flips both landed
2026-06-26	rev19	rev19: FIX B regressed, Core still 32/33	flat vs rev18, both expected flips did NOT happen
2026-06-26	rev8	rev8: route-config dump pins the suite 404	the answer is (ii), and we have the JSON
2026-06-25	rev7	the hostname-less routing fix did NOT flip traffic, rev7	the headline did NOT move
2026-06-25	rev5	first scored GATEWAY-HTTP, rev5	Delta: GATEWAY-HTTP went 6 PASS / 31 FAIL (rev3) → Setup-blocked / 0 (rev4) → 7 PASS / 30 FAIL (rev5).
2026-06-25	rev4	re-run rev4	Delta: GATEWAY-HTTP went from 6 PASS / 31 FAIL (37 run) in rev3 to Setup-blocked / 0 run in rev4 — a net regression in measurability, caused by a new cert-namespace resolution bug in …
2026-06-25	rev3	re-run rev3	Delta: GATEWAY-HTTP is 6 PASS / 31 FAIL — identical totals to rev2 — but the dominant failure category flipped from "Gateway has no address" (rev2, ~20 tests) to "HTTP→HTTPS redirect + …
2026-06-25	rev2	re-run rev2	Delta: GATEWAY-HTTP went from 0 tests run (aborted in Setup) → 37 tests run, 6 PASS / 31 FAIL.
2026-06-25	—	baseline run	GATEWAY-HTTP cannot run at all against aether's edge model as-is (an infra/addressing/namespace-scoping blocker, documented below — this negative result is itself the finding).