- From: Tianhao Liu via GitHub <noreply@w3.org>
- Date: Wed, 22 Jul 2026 20:58:53 +0000
- To: public-webrtc-logs@w3.org
I reviewed the current `supported-stats.https.html`. It still clears `remaining[type]` when a member is observed for either media kind, and `getAllStats()` only waits for either audio or video `remote-outbound-rtp`. A kind-aware structure seems mechanically straightforward: use `(type, kind)` buckets for stats that expose `kind`, retain type-only buckets for kindless stats, and include the kind in each subtest name. `media-source` can derive its members from `RTCAudioSourceStats` and `RTCVideoSourceStats`. The remaining blocker is the authoritative member mapping for the shared RTP dictionaries. Their WebIDL does not encode kind applicability. PR #774 added several normative `MUST NOT exist for audio/video` clauses, but #660 is still open, so I do not want to assume that every unannotated member is required for both kinds without confirmation. Would the intended policy be: 1. generate subtests only for spec-applicable `(type, kind, member)` triples, using an explicit exclusion table sourced from the normative `MUST NOT ... for <kind>` clauses; 2. generate no subtest for non-applicable triples; and 3. keep a missing applicable member as a normal failing subtest, leaving browser-specific expectations downstream, while reserving `PRECONDITION_FAILED` for the existing explicitly untestable cases? If so, I can prepare the WPT change in that form. -- GitHub Notification of comment by thliu21 Please view or discuss this issue at https://github.com/w3c/webrtc-stats/issues/707#issuecomment-5051479321 using your GitHub account -- Sent via github-notify-ml as configured in https://github.com/w3c/github-notify-ml-config
Received on Wednesday, 22 July 2026 20:58:54 UTC