Re: [webrtc-pc] Revert integration of URL parser for ICE servers from Recommendation text (#2997)

Running https://github.com/web-platform-tests/wpt/pull/47959 locally shows quite a number of cases where implementations fail due to not using the expected URL parser (both because they're too lax and too lenient).

It also highlights what I think are likely bugs in the spec:
* we don't forbid the `username@pasword:` construct; this gets rejected in Chromium, which I think is reasonable
* nowhere does the algorithm reject something like `stun:/example.net` or `stun:example.net/foo`, which I don't see any point in accepting

I'm pretty sure there would be valuable insights in detecting bugs further down the line in how server hostnames are passed to the ICE agent (which we could detect via `RTCIceCandidate.url`), but since that is currently not explicitly specified and since I'm not sure there is any infrastructure for that kind of ICE test, that's probably best left to later.

-- 
GitHub Notification of comment by dontcallmedom
Please view or discuss this issue at https://github.com/w3c/webrtc-pc/issues/2997#issuecomment-2328528755 using your GitHub account


-- 
Sent via github-notify-ml as configured in https://github.com/w3c/github-notify-ml-config

Received on Wednesday, 4 September 2024 10:38:40 UTC