- From: James Cloos <cloos@jhcloos.com>
- Date: Thu, 05 Dec 2013 14:42:51 -0500
- To: Martin Thomson <martin.thomson@gmail.com>
- Cc: HTTP Working Group <ietf-http-wg@w3.org>
>>>>> "MT" == Martin Thomson <martin.thomson@gmail.com> writes: MT> Using 2817 to upgrade to TLS is one of many options that has been MT> proposed, MT> though whether HTTP/2.0 appears in the Upgrade header or in MT> ALPN is very much undecided, Technically speaking, it is already there. Asking for tls using 2817 with the expectation to try to get 2.0 via alpn only requires that the server support rfc 2616, rfc 2817 and draft-ietf-tls-applayerprotoneg. (Or npn. ;^) Any server which supports 2817 and alpn would Just Work notwithstanding discussion here. But asking for both tls and 2.0 in a single upgrade request also should just work. You'd get whatever subset of the two the server supports, of course, and the client should proceed accordingly. But it is the only way optionally to upgrade to the common subset of { TLS HTTP/2.0 } shared by the client and server without wasting round trips. An example will answer questions such as "One Upgrade: header listing both desires vs multiple headers, one per desire", what the possible responces mean, et cetera. MT> particularly since we are discussing Alt-Svc, Alternate-Protocol, MT> and other options at the same time. Some of the threads have been too long-winded to get through in one sitting, but I haven't completely ignored any. And an example of how optimally to combine 3.2's Upgrade usage with the exsting 2817 Upgrade usage seems like the Right Thing To Do, even if other options may be added before publication as an rfc. -JimC -- James Cloos <cloos@jhcloos.com> OpenPGP: 1024D/ED7DAEA6
Received on Thursday, 5 December 2013 19:50:56 UTC