- From: Willy Tarreau <w@1wt.eu>
- Date: Tue, 7 Aug 2012 18:14:34 +0200
- To: Poul-Henning Kamp <phk@phk.freebsd.dk>
- Cc: Mark Nottingham <mnot@mnot.net>, James M Snell <jasnell@gmail.com>, Roberto Peon <grmocg@gmail.com>, ietf-http-wg@w3.org
Hi Poul-Henning, On Tue, Aug 07, 2012 at 03:58:34PM +0000, Poul-Henning Kamp wrote: > In message <E67EC141-85D9-4F64-920F-9F3EAB251DC7@mnot.net>, Mark Nottingham wri > tes: > > >I'd characterise expect/continue as being on the lighter borders of a > >grey area; it is used, but it does cause interop problems. > >[...] > >I'd be interested to hear comments from others, of course. > > Expect/continue should not be allowed in HTTP/2.0, it is a transport > flow-control mechanism and it does not work. No, it's not a transport flow-control mechanism, it goes further by letting the client send non-idempotent requests over established connections with the ability to reply them in case of breakage. By sending even a few bytes of data, you risk sending the whole request at once and being left in doubt when the connection breaks. It is important to be able not to send the whole request at once (at least stop before the last chunk). Right now the way it works is awkwards and could be improved or reworked, but there is a valid use of this mechanism even if it's sometimes quite painful. Willy
Received on Tuesday, 7 August 2012 16:15:03 UTC