Closed ThaUnknown closed 1 year ago
Can you elaborate on this?
Oh, noticed it's a draft. Nvmd.
Possible to update this to remove peers that aren't connected, instead of all of them? This is there basically to clean those up i/fwhen they timeout from the signalling side.
Possible to update this to remove peers that aren't connected, instead of all of them? This is there basically to clean those up i/fwhen they timeout from the signalling side.
that will remove peers who take a long time to connect, which I had happen
those probably should GC when they fail to connect?
idk how to manage either of those things tbh, I haven't looked at the methods of this lib too much
The server timeout is 2 minutes and the ICE timeout is typically like 30s. If the peers are taking long to connect I’m guessing they never were going to connect, not that they would eventually.
I’m up for revisiting this cleanup in the way you suggest after we merge the version that maintains active connections. I would need to set up a testing environment and so on to do a full test to make sure that doesn’t regress anything - but the simpler change where we don’t disconnect active ones seems obviously safe so let’s land that first.
removes peers removal from worker responses as its unreliable