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

[MikkelFJ <[e-mail]>]

I think it is pretty clear that later recieved lower numbered limits should be ignored or result in a protocol error - or ignore if the reciever cannot be bothered to check.

If it is explicitly required that limits are increased monotonically, it is protocol error, otherwise it is a ignore.

>From a quick read over the spec, I can't see that there are rules thay enforce that limits cannot be reduced sender size?

While it makes sense to transmit the largest limit on retransmission there are cases where this is not ideal because it can require synchronization. The same applies to packet numbers which must increase, but the more values that must be in sync, the harder it becomes to do parallel processing or to have separate responsibilities.

As to being helpful - I don't think the receiver should fix sender bugs - the most helpful thing is a hard protocol error, log output and a friendly email to the responsible dev team. If bugs starts to get worked around, ossificiation will ensue at scale. In this particular case it may not even be a bug, but a deliberate synchronizatin artefact.


-- 
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-409155257