Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Gorry's ECN rewrite #4059
Gorry's ECN rewrite #4059
Changes from 2 commits
d2ada7e
2a96678
245c51c
a00bb9b
0a22dfe
a6928b3
70e43d5
05e3f85
69f1f5f
1801df5
1a28ce1
4333941
96c8c6f
5c39ab2
0bcbe5c
fbfd077
File filter
Filter by extension
Conversations
Jump to
There are no files selected for viewing
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is a bad example as there's no practical use case where 0-RTT would be coalesced with 1-RTT as one would never have both keys at once. I understand the point of the example, but I think it's important to keep to practical examples.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That's true, except that this invalidates the other part of this paragraph... This needs more thought.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ok, happy to fix this in a separate PR if that makes sense?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think this is technically legal, so I'm suggesting a note about how it's extremely unlikely.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Setting ECN on 10 packets if you think they're going to be lost seems odd and I can't imagine anyone would do that. This text is such a departure from the existing text, I'm unsure how to fix it.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This isn't a departure from existing text -- see lines 3953-3956 in the old text.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So, I’m going to chime-in here. Indeed it was in the “original” text, which is why I included this. However, we now have several people (me included) who doubt this is yet correct, so here are my thoughts:
(1) I think it is important to be robust and describe how the protocol falls back in the corner case where the path drops IP packets with a non-zero ECN field - and we should very clearly say this.
(2) I think this can actually happen, but we should not make it sound like this often happens in general. I’m confident this is not “normal”.
(3) If we need to measure the path characteristic, using 10 packets seems a good test to me. However, as a fall back detection mechanism in a transport protocol this seems odd to stop using ECN after 10 packets when we expect that ECT is forwarded. Note: The text says {{ecn-ack}} says 10, the current text does not.
So.... This is a proposed change: We seem to be converging on clear text, would it be possible to remove the "10 packets" and revise this mechanism to say that the ECN counts need to increment, and pathological loss with no incrementing of counts is an indication that the sender should become “failed” for this connection? I think the text already virtually says this.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
"Pathological loss" isn't going to be good enough as a mechanism, and there's a rathole waiting there to define precisely what "pathological" means. We chose 10 packets based on the IW, and also because we wanted to start somewhere and then change the mechanism as we learned more through experience.
I suggest that we don't muddle this particular (large) PR with an additional, potentially-contentious discussion about a mechanism change.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I do like Jana's proposal to not muddle this particular (large) PR with this "issue". Would others agree to finalise the PR and then discuss this as a separate topic - afterwards?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think we discussed this all ready once. I'm fine with 10 packet as it's the initial window (but maybe that's what we should say), however, given the result is the same I would rather state it as enable ECT marking and if all packet of your first flight are lost validation fails and you disable. Describing this specifically as probably makes it just sounds more complicated that it is.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The github diff didn't line up well, hence my confusion. The way this is described still seems a bit unrealistic to me, but I can see this PR doesn't substantively change it, so I'm fine with that.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can fail? Isn't it required to fail?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is a reword of lines 3992-3996. I don't remember why this was written as a MAY earlier (as against a MUST, or just "fails"), but the new language is changing phrasing to "can" from "MAY". @martinthomson might remember why this isn't a failure requirement.