W3C home > Mailing lists > Public > public-webrtc-logs@w3.org > June 2020

Re: [webrtc-pc] Spec says scaleResolutionDownBy must default to 1, but browsers differ (#2535)

From: Florent Castelli via GitHub <sysbot+gh@w3.org>
Date: Tue, 02 Jun 2020 12:44:38 +0000
To: public-webrtc-logs@w3.org
Message-ID: <issue_comment.created-637517703-1591101877-sysbot+gh@w3.org>
I don't think we have any counters on this. I can look into adding that when I'm back to work in a few weeks and possibly drafting a deprecation plan with warnings for our users to set the right values.

Though, there is also another non-conformant behavior in play at the moment in Chrome where we disable layers automatically if the source is considered too small. It can be triggered if you set a low enough capture resolution or if any adaptation happens from a source that was at first big enough. 

The general problem with this is that there is little to no good signal the application can have to know when adaptation is happening to replicate a similar behavior. I tried disabling that feature and ran into issues from HW encoders refusing to work with small odd sizes, which you can get easily with a scaling factor of 4 and a low capture resolution.
If you have explicit scaleResolutionDownBy parameters and adaptation happens, you may want to react and adjust the layer count and scaling factors. It is hard to do that now, but combining the 2 non conformant behaviors, you get some sensible behavior.

It is tricky.

GitHub Notification of comment by Orphis
Please view or discuss this issue at https://github.com/w3c/webrtc-pc/issues/2535#issuecomment-637517703 using your GitHub account
Received on Tuesday, 2 June 2020 12:44:40 UTC

This archive was generated by hypermail 2.4.0 : Saturday, 6 May 2023 21:19:51 UTC