snownet: stagger sending of candidates based on connection state #4164
Assignees
Comments
thomaseizinger
changed the title
|
It might be better to still send them but delay acting on them. |
We !need to add them to the agent before sending so we know that they are actually unique and then the agent starts testing them so we have to delay adding them otherwise they fail because the other party isn't testing them at the same time. |
To start with, it is probably easier to do all relay candidates at once if we don't see a connection within a second or two. |
This was referenced Mar 22, 2024
|
Moving this to TODO because I think it will help with fixing #4290. |
|
I think we're pausing on this? Moving back to No Status |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Currently,
snownetimmediately seeds a connection with all available candidates when it is created and keeps adding new ones to it as they are discovered.With our current use of relays, that results in a lot of candidates that are being exchanged and lots of channels to be formed. In most cases, we should be able to hole-punch and create a connection within ~500ms - 1sec (signalling already subtracted).
Instead of sending all candidates immediately, we should start with host and srflx candidates. Then, if the connection is not established within 1sec, send over the relay candidates, potentially also one by one, i.e. a new candidate every ~100ms.
The text was updated successfully, but these errors were encountered: