Re: [quicwg/base-drafts] It's unclear if persistent congestion is a per-PN-space property (#3939)

[Marten Seemann <[e-mail]>]

> * When receiving an acknowledgement, mark a hole in the sentmap for the packet being acked.

I don't think this is trivial, quite the opposite actually. A "hole" would be extra state (note that just the absence of a packet at a certain PN does not constitute a hole, since you could've skipped that PN to mitigate the optimistic ACK attack), and that additional state has to be cleaned up.

In my implementation, when a packet is acknowledged, I know I can delete it and get rid of all state associated with it. Deviating from this, just to be able to do do persistent congestion detection across packet number spaces, seems like a lot of hassle to me.
As @janaiyengar described, any packet loss during the handshake will half the congestion window, and with #3943, it will even reduce the congestion window to half the bytes in flight (which will commonly be less than the cwnd during the handshake). Simple Reno-style congestion control will therefore lead to behavior that is very similar to actually doing persistent congestion detection across packet number spaces, without all the additional complexity.

-- 
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/3939#issuecomment-662811684