As far as I'm concerned, you're welcome to do whatever you want with a CONNECT tunnel. If you want to layer WebSockets on top of it, go for it. There are lots of other considerations with regard to that (see https://groups.google.com/forum/#!topic/spdy-dev/rwOh5dH4ibU for some previous discussion). I think this is probably out of scope for HTTP/2 discussion, at least at this early stage. On Tue, Aug 13, 2013 at 11:06 PM, Ilari Liusvaara < ilari.liusvaara@elisanet.fi> wrote: > On Tue, Aug 13, 2013 at 09:40:18PM +0100, Martin Thomson wrote: > > On 13 August 2013 20:24, Ilari Liusvaara <ilari.liusvaara@elisanet.fi> > wrote: > > > END_STREAM <----> Half-close > > > > That reminds me. Treating this as an approximation for a TCP > > connection isn't perfect if you permit the stream to remain half > > closed for a significant period. I think that you have to treat > > END_STREAM as being equivalent to a TCP FIN. > > At least for some things running over TCP one has to be able to half- > close the connection and precive peer's half-close or it just plain > won't work. > > > > As note, this sort of stuff could also be applicable to upgrade > > > (internally emulate TCP connection)? > > > > I see no reason to support upgrade. We're providing plenty of other > > hooks for upgrading prior to commencing HTTP/2.0. You can use ALPN or > > HTTP/1.1 Upgrade. > > I don't mean X -> HTTP/2.0 upgrade, but upgrading a stream from HTTP/2.0 > to something else. > > E.g. Full-blown websockets implementation (as opposed to potentially > simplified one built into HTTP/2.0). > > -Ilari >
Received on Tuesday, 13 August 2013 21:22:36 UTC