Re: HTTP/2 response completed before its request

Note that it feels like the scenario is very contrived.

If either side (including the client) isn't reading, it cannot receive a
RST_STREAM, nor WINDOW_UPDATE or anything else.
Such a client or server would be almost guaranteed to fail.

So, essentially, the bug here would be on the side that isn't reading,
regardless of the endpoint being client or server.

Any side which ceases reading/examining received bytes can cause the
session to enter into a deadlock state.

-=R


On Tue, Jul 1, 2014 at 1:31 PM, Poul-Henning Kamp <phk@phk.freebsd.dk>
wrote:

> In message <
> CA+pLO_gzKNhCXTvf7Uwkcky2uEWiUYJNnH-ArxuJOSyC-ic_Eg@mail.gmail.com>, Jeff
> Pinner writes:
>
> >The client can't send END_STREAM to close the response early since
> >that would not differentiate an incomplete payload body from a
> >complete one. It would need to send RST to indicate early termination
> >of the payload.
>
> Given that the server already said "Screw the payload", I can't see
> what difference that could possibly make to anybody, but if RST
> makes you more happy, then send RST...
>
> The worst that can happen is that you pollute error-log on the
> server.
>
> --
> Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
> phk@FreeBSD.ORG         | TCP/IP since RFC 956
> FreeBSD committer       | BSD since 4.3-tahoe
> Never attribute to malice what can adequately be explained by incompetence.
>
>