Re: END_SEGMENT? (#397)

On 18 June 2014 17:46, Yutaka Hirano <yhirano@google.com> wrote:
> I'm interested in WebSocket over HTTP/2 and it uses END_SEGMENT.
> The problem of specifying END_SEGMENT in WebSocket over HTTP/2 not in HTTP/2
> is, WebSocket over HTTP/2 can't be used with intermediaries that merely
> understand HTTP/2.

I'm less concerned about this particular problem, especially now that
we've effectively opened the door to extension.  Websockets is always
going to be dependent - at least a little - on the next hop
understanding something of the request.  We've talked in the past
about a new ALPN token for that, but it might be as simple now as
SETTINGS_ENABLE_WEBSOCKETS.

It would be relatively easy to port the necessary flags to a
websockets draft if we decided to remove them from HTTP/2.

I do think that we should be looking more seriously at your draft.  In
light of recent changes, but more because we agreed to do this, and I
think that you have the most fully developed proposal that we've
deferred due to increased attention on getting HTTP/2 done.

Received on Thursday, 19 June 2014 02:21:39 UTC