Re: RST_STREAM and FINAL flag

On Thu, Apr 25, 2013 at 12:12 PM, Martin Thomson
<martin.thomson@gmail.com> wrote:
> James raised this:
>
> https://github.com/http2/http2-spec/issues/77
>
> "... what is not explicitly stated anywhere in the draft is that the
> FINAL flag for the RST_STREAM frame MUST be set. Likewise with
> GOAWAY..."
>
> I don't think that it matters what the FINAL flag says for RST_STREAM
> - processing the frame closes the stream in totality, so whether the
> stream gets half-closed in the process is largely irrelevant.  Same
> for GOAWAY.

+1 ... for completeness, the definition of the RST_STREAM and GOAWAY
frames can say that the FINAL flag is to be ignored in all cases
because the frames themselves are terminal in nature.

>
> Also, "what happens if an endpoint sends a RST_STREAM with stream
> identifier #0 ? "
>
> My take: this is very bad.  Don't do that, but if you do see it, then
> it's a PROTOCOL_ERROR.
>

Agreed.

- James

Received on Thursday, 25 April 2013 19:18:02 UTC