Closed gorryfair closed 6 months ago
Raffaello: Safe Retreat ought to be conservative and also drain the pipe, but we can learn from the measured pipe by observing the volume of data ack'ed in validating. We will look at updating this part of the spec in a PR
Kazuho: What happens when there is tail-loss at the end of the unvalidated phase, - we need to avoid that we set an unduly low CWND.