- From: Willy Tarreau <w@1wt.eu>
- Date: Thu, 29 May 2014 08:24:17 +0200
- To: Eliot Lear <lear@cisco.com>
- Cc: James M Snell <jasnell@gmail.com>, Nicholas Hurley <hurley@todesschaf.org>, "Richard Wheeldon (rwheeldo)" <rwheeldo@cisco.com>, HTTP Working Group <ietf-http-wg@w3.org>, Michael Sweet <msweet@apple.com>, Martin Thomson <martin.thomson@gmail.com>
On Wed, May 28, 2014 at 05:57:07PM +0200, Eliot Lear wrote: > > On 5/28/14, 5:52 PM, James M Snell wrote: > > > > So are we getting back to the "they can just keep using http/1.1" > > argument or do I misunderstand your point? > > > > Yes. That's not to say people shouldn't adapt HTTP/2 to other uses. > Just that HTTP/2 needn't adapt to *all* uses. And perhaps the onus > should be on those who don't want to do compression to indicate that > HTTP/2 to them is still useful. If there are enough of them, this is an > easily solved problem: just use another identifier of HTTP/2 in ALPN > that indicates no compression. h2c is taken. My opinion has always been that HTTP/2 should be able to replace HTTP/1 in all situations, even if not optimally in some cases. Having to maintain dual-stack in many products will remain a security challenge and at some point client developers will start to emit warnings when using 1.1 : "Warning: the site you are trying to visit uses an obsolete version of HTTP which can cause your system to malfunction, do you want to continue ?" And just because of that, all commonly used devices will have to support 2.0 even with the minimal feature set. So I'd like HTTP/2 to be able to replace HTTP/1 in all use cases even if that's not optimal nor justified by technical reasons. Willy
Received on Thursday, 29 May 2014 06:24:50 UTC