Re: [quicwg/base-drafts] all HTTP/3 protocol violations should lead to connection errors (#2510)

[Kazuho Oku <[e-mail]>]

@dtikhonov 
>> An endpoint MUST NOT send a header list that exceeds the peer-advised limit.
> 
> How can this be justified? This is a departure from RFC 7540, which states in two places that this setting is only advisory.

Yeah. My read of HTTP/3 is that it is already different from HTTP/2 this respect. It is no more clarifies as "advisory," no there is a mention that a client can exceed the limit (a text that exists in RFC 7540).

Regarding the ideal behavior, I prefer being strict here, because it would give us the most predictive behavior at a minimal cost (it's trivial to check the size, isn't it?). Think it from an end-user perspective. Without the restriction, we might see one user-agent telling you that the request is too large, and the other processing the same request. That'd be confusing.

-- 
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/pull/2510#issuecomment-471710008