Re: [quicwg/base-drafts] Fix to ECN section regarding validation (#2113)

[Magnus Westerlund <[e-mail]>]

Benjamin Saunders did point out that if one have ACKs with none overlapping ACK ranges and they get reordered there is still an issue with performing the comparison in some cases. Example:

Receiver sent: 
ACK:1  PN:1,2,3,4 ECT(0)=4, ECT(1)=0, CE=0
ACK:2  PN:5,6,7,8 ECT(0)=8, ECT(1)=0, CE=0
ACK:3  PN:9,10,11 ECT(0)=11, ECT(1)=0, CE=0

Sender receives:
ACK:1  PN:1,2,3,4 ECT(0)=4, ECT(1)=0, CE=0
ACK:3  PN:9,10,11 ECT(0)=11, ECT(1)=0, CE=0
ACK:2  PN:5,6,7,8 ECT(0)=8, ECT(1)=0, CE=0

Local Reference Counters: 0,0,0
1: Comparison ECT(0)_diff=4, Newly ACKed=4 (with ECT(0)) OK
Local Reference Counters: 4,0,0
3: Verification ECT(0)_diff=7, Newly ACKEd=3
Local Reference Counters: 7,0,0 (Based on ACKed packets)
Local Reference Counters: 11,0,0 (Latest Received ACK)

2: Verification ECT(0)_diff=1 or -3 (Based on ACKed, vs Latest Received).
Newly ACKed=4
Comparison fails if run.  

So the question is how to deal with this. From my perspective there are multiple implementation strategies that could solve this. So my question is should we mandate any particular or simply try to illustrate the issue. I think the most straight forward and which I don't so far sees any flaws with are:

- Don't perform verification with ACK frames that on arrivial is not highest PN seen. 

Opinions?

-- 
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/2113#issuecomment-446658001