Re: [quicwg/base-drafts] QUIC's Initial Congestion Window specification is incorrect (#3997)

[ianswett <[e-mail]>]

Recovery says "QUIC begins every connection in slow start with the congestion window set to an initial value."

There's no mention of updating the initial congestion window, so at the beginning of the connection is currently the only time it's set.

QUIC advises setting do not fragment, and UDP typically gets dropped instead of fragmented in my experience even without the flag set, so I believe that simplifies the situation.

Given that, along the lines of (iii), I can imagine one potential issue is if the initial window is set to a large value based on some past information and then the MTU is decreased in order to allow the handshake to complete.  Currently there's nothing saying that the IW should be re-calculated if the max packet size decreases during the handshake, and adding one would be along the lines of your second suggested option.

But I'm also fine with removing the twice max datagram size clause, but that seems like a larger departure from existing RFCs, which makes me hesitant.

https://github.com/quicwg/base-drafts/blob/master/draft-ietf-quic-recovery.md#initial-and-minimum-congestion-window-initial-cwnd

-- 
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/3997#issuecomment-673695661