- From: Yoav Nir <ynir@checkpoint.com>
- Date: Mon, 22 Jul 2013 19:44:31 +0000
- To: Ilari Liusvaara <ilari.liusvaara@elisanet.fi>
- CC: Zhong Yu <zhong.j.yu@gmail.com>, Mike Belshe <mike@belshe.com>, "HTTP Working Group" <ietf-http-wg@w3.org>
On Jul 22, 2013, at 10:24 PM, Ilari Liusvaara <ilari.liusvaara@elisanet.fi> wrote: > On Mon, Jul 22, 2013 at 11:47:48AM -0500, Zhong Yu wrote: >> Suppose a TLS connection is established without ALPN. Then an HTTP/1.1 >> request is sent over with Upgrade: HTTP/2.0. How should the server >> respond? > <snip> > > There's also second abnormal case: > TLS handshake without ALPN, but the application data starts with > HTTP/2.0 magic… I'm guessing that if either the TLS application data or the TCP connection in HTTP starts with the magic, it's HTTP/2.0. Otherwise, why would we need the magic^H^H^H^HConnection Header? >> Though "Upgrade" mechanism is less ideal than ALPN, since the server >> must support it anyway on TCP connections, I don't see why we should >> forbid it on TLS connections. > > What about servers that are not willing to implement upgrade (it is fair > amount of complexity)? They'd have to stick to HTTP/1 for now, because no sane client start off with HTTP/2 for now. > I.e. is there path to totally obsolete HTTP/1.1 in the far future? Not by having servers or client that refuse to negotiate the version. It's also possible that by the time we can assume that we don't need HTTP/1 any more, we may have to contend with HTTP/3. Yoav
Received on Monday, 22 July 2013 19:45:17 UTC