Re: [quicwg/base-drafts] Pacing when under-utilizing the Congestion Window (#2686)

[Jana Iyengar <[e-mail]>]

Gorry: Yes, I agree that micro-bursts ought to be smaller than IW. That's sensible, I think, but we really don't have a best practice for micro-bursts that we can point to (I just remembered that I had written a [draft](https://tools.ietf.org/html/draft-iyengar-burst-mitigation-01) with mallman 14 years ago on trying to specify micro-burst behavior!).

Yes, RFC 4960 uses a simple micro-burst of 4 segments, but that was based on RFC 3390, which used to limit the initial window to at most 4 segments. That's changed now, but I'd be wary of reflecting that change onto micro-bursts, because the analysis is quite different.

So while I agree that largest(max_burst) <= IW, I think talking about max_burst adds something new to what the draft is currently talking about. The draft is simply using IW for both connection start-up and noting that as a known upper bound for bursts. That is not the same as specifying max_burst however.

I think we ought to say something for senders coming out of an app limited period. (Perhaps we should say that bursting SHOULD be limited in the absence of pacing and add a note to various past max_burst limits without specifying a new one?)

I'm not keen to specify a max_burst here (but I'm not sure if that's a tenable position).

-- 
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/2686#issuecomment-493289281