First off, thanks so much for all the great work on the Jitsi project.
It's a very awesome suite of voice/video tools.
In the course of trying to do some unusual things with Jitsi-Videobridge, I
found a bug in the ice4j code with regards to conflict resolution.
I posted a bug with a patch for it on the ice4j project at:
Basically, it looks like we're doing tie breaker comparison incorrectly in
ice4j, causing role conflict resolution to fail sometimes in ways that
would conflict with the specifications.
Another thing I added in it was a way to override the tieBreaker value so
that Jitsi-Videobridge could always win. Cases arise where
Jitsi-Videobridge does not operate properly if it is in the ICE-CONTROLLED
role in an aggressive nomination environment such as used by Chrome or (I
think) more recent Firefox.
Please let me know what you think.