Re: [quicwg/base-drafts] compensation of ack_delay is fragile against errors (#2060)

[janaiyengar <[e-mail]>]

Responding to the most recent couple of comments first: SRTT does not include ack_delay, which is why we add max_ack_delay back into the RTO computation. So, even in the case of ack loss, the SRTT *should* be the same... the reported ack_delay might keep increasing.

There is the question of what to do if ack_delay gets larger than max_ack_delay, which is why I had argued that should prefer observed ack_delays over explicitly stated ones (in an earlier thread.) I think we had discussed monitoring ack_delays and using that value over time, but left it for later...
ideally, we'd use an EWMA or some such to track actual ack_delays to see if they're increasing, and add that into the RTO computation.


-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/2060#issuecomment-443075810