Re: [webrtc-pc] Should garbage collecting RTCPeerConnection be observable? (#3090)

I think the internet is doing fine without GC side effects. Any virtue that relies on a garbage collector seems a poor one to me. 
It's basically saying that JS apps shouldn't worry about closing their connections because the GC will do it for them.

This principle collides with another: [§ 5.3. Don’t expose garbage collection](https://w3ctag.github.io/design-principles/#js-gc) which says good GC is undetectable, a browser optimization entirely _outside_ specs, ideally without any observable behavior.

I think @karlt in [bug 1805320 (comment 7)](https://bugzilla.mozilla.org/show_bug.cgi?id=1805320#c7) says it well:
- _"If GC were to cause observable effects then it would be collecting something that is not garbage."_
- _"The spec is clarifying some cases where GC must not occur. It is not intending to specify the precise criteria for when GC may occur. GC may occur whenever it would not be observable, and so the spec does not need to describe GC"_

His second point seems relevant to our spec and how we read https://w3c.github.io/webrtc-pc/#garbage-collection

I think the only part that might not have come up before is: does observable behavior extends to a remote peer? The example above show it's trivial in WebRTC to set up a local loop connection, so I would argue it does.

-- 
GitHub Notification of comment by jan-ivar
Please view or discuss this issue at https://github.com/w3c/webrtc-pc/issues/3090#issuecomment-3967166947 using your GitHub account


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

Received on Thursday, 26 February 2026 14:58:40 UTC