- From: Stefan Eissing <stefan.eissing@greenbytes.de>
- Date: Wed, 20 Nov 2013 10:55:14 +0100
- To: Yoav Nir <synp71@live.com>
- Cc: Willy Tarreau <w@1wt.eu>, HTTP Working Group <ietf-http-wg@w3.org>
Am 20.11.2013 um 09:23 schrieb Yoav Nir <synp71@live.com>: > A new scheme means changing stuff at other layers (HTML, links in emails, lots of other places). We could do something like this: > > * http://www.example.com - soon: happy eyeballs. later: attempts HTTP/2 and fallback if that fails. > * http://www.example.com:80 - HTTP/1 > * http://www.example.com:100 - attempt HTTP/2 and fallback if that fails. > * http://www.example.com:xxxxx - soon: begin with HTTP/1, upgrade if you get an alt-svc for the same port. later on: attempt HTTP/2 and fallback if that fails. > > Not pretty, but getting people to migrate to giving out http2:// urls is something I don't think is going to happen. I think, sooner or later, all "http*://hostname/path" need to point to the same resource. "http" becomes then an umbrella scheme for http1/https/http2/futurehttp. (and "https" the umbrella for all secure variants of http) It is awkward that services like google need to redirect URIs all the time, just to encrypt the traffic. And they will need to continue doing so when the next, even better, protocol variation comes. It believe it's naive to assume that port 443 will solve all future protocol upgrade scenarios. <green/>bytes GmbH Hafenweg 16, 48155 Münster, Germany Phone: +49 251 2807760. Amtsgericht Münster: HRB5782
Received on Wednesday, 20 November 2013 09:55:37 UTC