Improving WebRTC video conferencing

[Jim Garrett]

I've been trying out various Jitsi Meet instances (including the
FSF's) and BigBlueButton (BBB) instances and it seems that both could
use some improvement.  I think the issue concerns WebRTC but I don't
know how to move forward.  I think we're currently failing to match
Zoom in quality and we need to figure out how to address the issues
below.

WebRTC ("Web real-time communications") is a standard implemented in
many browsers for interacting with the camera, microphone, and also
handling signal processing and encoding/decoding.  I'm not an
expert, but it seems that Jitsi Meet and BBB are not really handling
the low-level stuff.  So when things don't work, or don't work well
enough, how does one fix it?  How do we even diagnose it, when we have
a variety of browsers and browser versions?

Interestingly, the issues I've had in meetings mostly concern audio,
not video:
- Some participants' sound-cancellation worked poorly, so they get an
  echo when they talk.  (Cancellation of local audio is a part of the
  WebRTC standard.)
- Some participants' audio corrupted into a high-pitched unintelligible
  form resembling Animal Crossing creatures.  If you don't happen to
  have family members using this non-free game software, think Charlie
  Brown adults' speech sped up and raised about 3 octaves.

These issues are intermittent, affecting some people but not others.
Quality software behavior requires that if we can't fix the problem, 
at least we identify conditions for which the software will work
successfully.  Most likely this comes down to browser and version
(hence the WebRTC implementation).  Maybe this is already the case and
that browser is Chrome/Chromium; if so, this poses another problem
because Chromium is not fully Free.

While audio has been problematic, video has worked pretty well.
Occasionally someone has failed to get their video to work.

Why am I posting this here rather than going directly to the
projects themselves?  I'm not a developer and would
appreciate the ideas of this community before I do that.

The best way I can think of to move forward is for these systems to log
users' browsers and versions, and query them at the end of the session
to rate and comment on their experience.  This would provide data that
might indicate specific browsers and versions--or interactions among
them!--that cause problems.  Then developers can start to identify if
in fact issues are specific to browser versions.

There's a question of how we can gather user data for important
software improvement efforts while respecting privacy (if in fact data
logging is the way to address these issues).  Any thoughts on how to
manage this?

Thanks for your thoughts.

-Jim Garrett