Re: [quicwg/base-drafts] Improve ACK_ECN frame encoding (e.g., use bit-vector) (#1439)

[Kazuho Oku <[e-mail]>]

A straw-man that addresses the issues is to use a bit-vector for marking CE packets.

Instead of carrying three counters, an ACK_ECN frame will carry a bit-vector that indicates which PNs in the ACK were CE-marked.

For example, when 3 packets (PN=8,3,2) are acknowledged, the bit vector will contain three bits that indicate whether the packet identified by the PN was CE-marked. A bit vector with a value 100 will indicate that PN=8 was CE-marked.

The straw-man addresses all of the three issues.

Re space efficiency of the encoding, the proposed approach consumes one bit per packet, whereas #1372 consumes 3 to 17 octets<sup>1</sup> even when only one packet is being acknowledged.

Duplicate detection becomes the responsibility of the sender of the packet rather than that of the receiver. It also becomes reliable.

While processing an acknowledgement, an endpoint consults it's inflight packet map to see which packets have been newly acknowledged. With the proposed encoding, the endpoint can see if the CE bit has been set as part of that process, and only utilize the CE bits associated to a newly acknowledged packet.

The downside of the straw-man is that ACK_ECN frame does not indicate how many packets were ECN(1) marked. If we need to communicate that, we could for either retain the counter for ECN(1) marked packets, or use a different encoding in the bit vector (i.e. 0 - ECN(0), 10 - ECN(1), 11 - ECN-CE).
 
[1] It is assumed that the number of ECN(1) marked packets would remain zero.


-- 
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/1439#issuecomment-396433945