Re: [quicwg/base-drafts] Retaining the largest received packet number (#3541)

[ianswett <[e-mail]>]

I'd like to suggest tracking largest acknowledged from forcing all ACK frames to contain the largest acknowledged and have that value never decrease.

Given you need to track largest acknowledged locally(ie: not only in the in-flight ACK frame), I don't think there are any problems there.

I have two cases in mind where the largest acked could validly decrease:
1) A receiver that tracks a limited number of ACK blocks and the oldest block happens to be the one containing largest acked.
2) If an application was reordering tolerant(ie: from https://tools.ietf.org/html/draft-iyengar-quic-delayed-ack-00) then two different senders on a single 'connection' may be sending packets with disparate packet numbers, causing what appears to be very large scale reordering.  In this case, it'd be valuable to discard the older ACK ranges even if they contained the largest acked.

Maybe the first is an edge case we don't care about in a use case we aren't optimizing for and the latter is solvable with careful issuing of packet number ranges.  Even so, I'm not sure what the motivation is for this change?  I guess the ECN case could be, but using largest acked as a proxy for retransmitted ACKs is a very bad design in my opinion.

-- 
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/3541#issuecomment-604984777