To: new-httpd@hyperreal.com
Subject: Re: Blaze?? 
In-reply-to: Your message of "Tue, 03 Dec 1996 15:17:39 GMT."
             <199612031517.PAA00295@aaaaaaaa.demon.co.uk>
Date: Tue, 03 Dec 1996 10:29:13 -0800
From: "Roy T. Fielding" <fielding@liege.ICS.UCI.EDU>
Message-ID: <9612031029.aa18804@paris.ics.uci.edu>
Sender: new-httpd-owner@apache.org
Precedence: bulk
Reply-To: new-httpd@hyperreal.com

> More on this Blaze thing.  Suppose you had two xSpeed aware poxies
> back to back.  They could send requests to each other with 'Accept:
> xSpeed/compressed' (or some such MIME), and once satisfied could
> then do the compression/encapsulation dance to send responses.

They don't even need that.  See the Upgrade header in HTTP/1.1.

> Mmm, besides, once two proxies have identified that they understand
> the same transfer protocol they can just open a UDP line and send
> their information down that, provided it doesn't abuse HTTP's
> various safeguards.
> 
> Roy F, is this just hokum?  Is there already a sensible mechanism
> for speeding up transfers between Proxies that doesn't go against
> the letter of the HTTP law?

The idea was for HTTP/2.0 to be a multiplexing, tokenized protocol.
The only thing it lacked was an persistent person to do the work of defining
it, as I did for HTTP/1.x.  The Upgrade header is the graceful deployment
mechanism.  Henrik already has a working version of this for Jigsaw.
Using a proprietary protocol for this kind of thing would be really stupid.

.....Roy