W3C home > Mailing lists > Public > ietf-http-wg@w3.org > October to December 2012

Re: #385: HTTP2 Upgrade / Negotiation

From: Adrien W. de Croy <adrien@qbik.com>
Date: Thu, 25 Oct 2012 21:39:52 +0000
To: "Willy Tarreau" <w@1wt.eu>, "Patrick McManus" <pmcmanus@mozilla.com>
Cc: "Yoav Nir" <ynir@checkpoint.com>, "Gabriel Montenegro" <Gabriel.Montenegro@microsoft.com>, "Mark Nottingham" <mnot@mnot.net>, "Amos Jeffries" <squid3@treenet.co.nz>, "ietf-http-wg@w3.org" <ietf-http-wg@w3.org>
Message-Id: <em482594e0-a9e2-4b58-8abe-0e03525b0750@bombed>

------ Original Message ------
From: "Willy Tarreau" <w@1wt.eu>
>Thus I would not be surprized to see a success rate close to 100% with the
>following sequence to sites accepting both ports when HTTP/2 is released :
> 1) HTTP Upgrade on port 80
> 2) fallback to TLS on port 443
>
>
I think 2 could introduce a significant delay.  What proportion of 
websites support TLS on 443?

Or is this purely to get past intermediaries to a site you already know 
supports 2.0?  Or how do you know already that the site is available on 
443, and if someone clicked a http:// URL, is it valid to make a https 
connection?  Sometimes it's a different site on the different port.

In fact for that reason alone, you can't change the port that the URI 
specified from 80 to 443.  You can end up getting the wrong site.

I foresee a bunch of problems where 1 fails due to intercepting proxy 
not understanding Upgrade, and 2 fails because the site is http only on 
80 only.

Or did I misunderstand what you're getting at?

Regards

Adrien
>
>And when both fail, clearly 1.1 is the only way to go.
>
>Regards,
>Willy
>
>
>
>
Received on Thursday, 25 October 2012 21:40:15 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 25 October 2012 21:40:17 GMT