Firefox still has an open bug regarding agressive ICE nomination with an ICE-lite peer Bugzilla. This means that ICE will fail for certain scenarios with Firefox. My question is this, would the maintainers of Mediasoup accept going against the RFC on this to support Firefox if a PR were to come their way? If so, would there be any caveats to this (e.g. behind a config option or compile flag)?
If that would be possible, my proposal would be to ignore USE-CANDIDATE and just use the last valid tuple instead. This would mean that ICE state would be somewhat wrong, in that you would have no clear distinction between the CONNECTED and COMPLETED state.
Ok. Can you please create a feature request in GH and expose the solution you have in mind? Some time ago I considered the option of sending a binding request from the server every time a binding request is received but I don’t know if it makes sense.
I want to understand all the options first and know in advance why a specific proposal would solve the problem. I’ll participate in the GH issue soon, it’s just that there are other pending things (PRs and so on) to manage.