Re: [quicwg/base-drafts] Proposal for adding ECN support to QUIC. (#1372)

[Kazuho Oku <[e-mail]>]

kazuho commented on this pull request.



> +
+The receiver side should implement three 64-bit counters that are copied to the
+ECN block when an ACK_ECN frame is generated:
+
+ECT_0:
+: Initial value = 0, incremented when a packet marked ECT(0) is received
+
+ECT_1:
+
+: Initial value = 0, incremented when a packet marked ECT(1) is received
+
+CE:
+
+: Initial value = 0, incremented when a packet marked CE is received
+
+Reception of duplicate packets SHOULD NOT increment the counters.

Can we count the packet (for all three counters) *before* detecting duplicates instead?

IMO, theoretically a receiver should send back all the information it has obtained. Ignoring the CE flag in duplicate packets goes against that principle. 

Counting packets after detecting duplicates makes sense only when we can assume that duplication happens near to the receiver than where ECN is applied. I do not think that we have such an assumption.

In practice, the recovery draft with the proposed changes only uses the CE counter to detect if _any_ congestion was reported by the ACK_ECN frame (it's essentially a binary flag). Considering that, I do not think that changing the moment of counting the packets causes any harm.

And lastly but importantly, by changing the rule as such, we can get rid of the requirement to detect duplicates.

See also: https://github.com/quicwg/base-drafts/issues/1405#issuecomment-393743048

-- 
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/1372#pullrequestreview-125194246