- From: Poul-Henning Kamp <phk@phk.freebsd.dk>
- Date: Wed, 20 Nov 2013 08:23:38 +0000
- To: Willy Tarreau <w@1wt.eu>
- cc: Adrien de Croy <adrien@qbik.com>, James M Snell <jasnell@gmail.com>, Mark Nottingham <mnot@mnot.net>, Bjoern Hoehrmann <derhoermi@gmx.net>, HTTP Working Group <ietf-http-wg@w3.org>
In message <20131120072826.GH22150@1wt.eu>, Willy Tarreau writes: >On Wed, Nov 20, 2013 at 04:08:49AM +0000, Adrien de Croy wrote: >> >> my 2c is that http/2.0 (TLS or not) is enough of a departure from http, >> that trying to put plaintext http/2.0 over port 80 will just be an >> impossible nightmare. > >Changing the port will require to change the scheme as well otherwise >it will end up being even worse. For example, *right now* over the net >and even much more in corporate networks, you have many applications >running on non-80 ports. So when the browser will have to connect to >"http://foo.bar.tld:8080/", what version will it use ? Long time ago I argued that we should look into "no-RTT upgrade", ie a scheme where the first byte sent from client to server on HTTP/2 would allow the server to decide which protocol it was. Together with an "Also:" or alt-svc header which announces which protocols are supported by the server, this could work seamlessly: if port == 100 http/2 elif we've seen Also: header announcing http/2 http/2 else http/1 This idea may be worth revisiting. -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 phk@FreeBSD.ORG | TCP/IP since RFC 956 FreeBSD committer | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence.
Received on Wednesday, 20 November 2013 08:24:00 UTC