Re: [quicwg/base-drafts] Persistent congestion pseudocode to match text (#4010)

[Kazu Yamamoto <[e-mail]>]

@martinthomson I have implemented this in Haskell quic.

1) With this pseudocode, *global* sequence number (unique and contiguous across the packet number space) is necessary to find the largest contiguity. In Haskell implementation, no additional field is necessary because packet numbers are unique globally.

2) It seems to me that `pc_lost` is over engineering if it is used only to detect persistent congestion. However, this structure can be also used to prevent spurious retransmission due to reordering, pointed out @kazuho. For example, if an endpoint receives ACK 1,3 and then ACK 2, the current pseudo code interprets that packet 2 is lost. If we use `pc_lost` as lost packet candicates and delay the decision, this issue can be resolved. So, we can justify the introduction of `pc_lost`. :-)

3) `pc_lost` consumes much resource because it is cleared only when persistent congestion happens. But, if we use `pc_lost` as lost packet candidates, it should be reduced when retransmission. Since the pseudocode does not describe how retransmission is implemented, we can left how to reduce `pc_lost` as implementation dependent and need not to implement it in this pseudocode.

-- 
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/pull/4010#issuecomment-678469958