Re: [quicwg/base-drafts] are flow control frames really idempotent? (#1612)

[Dmitri Tikhonov <[e-mail]>]

@siyengar writes:
> a receiver who wants to help peers prevent flow control bugs from happening

One could detect _potential_ bugs in the peer in many other cases.  I vote against mandating such error detection at the protocol level.

@kazuho writes:
> some existing QUIC implementations simply retransmit a packet (by re-encrypting it using a new PN) when it is deemed lost. I think that such implementations should be allowed to exist.

This is true. This is what the LiteSpeed implementation does with GQUIC's _WINDOW_UPDATE_ frames. The smaller values are [ignored](https://github.com/litespeedtech/lsquic-client/blob/04468d215dc184ae64e7aa1d0e940da50c724e2f/src/liblsquic/lsquic_stream.c#L1855). It is more expedient (and is cheaper) to discard lower values on receipt rather than update or remove the old values on retransmit.

Yes, let us exist!

-- 
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/1612#issuecomment-409203779