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

[ianswett <[e-mail]>]

Yes, those are the two issues, and they're somewhat separate.

For 1, I'm concerned we could ignore samples for a long time under some circumstances.  If the RTT is 25ms and a connection is always sending 1 packet and hitting the delayed ack timer, then the min_rtt will be 50ms, not 25ms.  And I believe ignoring SRTT samples that are less than min_rtt would result in ignoring every SRTT sample until an immediate ack was sent.  Admittedly, we say to ACK crypto packets immediately, but if one used a 1ms timer as 'immediate', I think the same problem results?

For 2, I think putting it into SRTT is the natural thing.  Adding another EWMA doesn't seem worthwhile, and in most cases it is very similar to incorporating it into SRTT.

-- 
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-443213389