Re: END_SEGMENT? (#397)

>
> > I don't want to require all intermediaries to understand WS over HTTP/2.
> If this is the requirement, then you'll need to retain END_SEGMENT.

What I really need is that an intermediary which don't understand WS over
HTTP/2 don't
do unnecessary "intelligent" things.
Specifically, I don't want an intermediary to drop a WS frame or to buffer
WS frames over
a message boundary without understanding WS over HTTP/2.

If you could state that an intermediary MUST NOT discard or buffer an
unknown or unsupported frames,
it would serve (without END_SEGMENT).
Sorry I didn't follow the "Extending HTTP/2" discussion, so please let me
know if the above statement is pointless.

I'm not convinced that this is right though.  I'd rather have an
> explicit acknowledgement that my protocol is understood by the
> entities that are participating in it.

That requires an additional statement on the HTTP/2 spec.
Though I will be happier if WS over HTTP/2 can work with intermediaries
that don't understand WS over HTTP/2,
WS over HTTP/2 can work with the "explicit acknowledgement" mechanism.

Thanks,



On Fri, Jun 20, 2014 at 12:44 AM, Martin Thomson <martin.thomson@gmail.com>
wrote:

> On 19 June 2014 00:51, Yutaka Hirano <yhirano@google.com> wrote:
> > I don't want to require all intermediaries to understand WS over HTTP/2.
>
> If this is the requirement, then you'll need to retain END_SEGMENT.
>
> I'm not convinced that this is right though.  I'd rather have an
> explicit acknowledgement that my protocol is understood by the
> entities that are participating in it.
>
> Imagine what would happen if an intermediary decided to cache the
> response to your websocket request...
>