W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > May to August 1996

Re: proposed HTTP changes for charset

From: Francois Yergeau <yergeau@alis.ca>
Date: Fri, 5 Jul 1996 17:37:58 -0500
Message-Id: <199607052140.RAA03403@genstar.alis.ca>
To: Ted Hardie <hardie@merlot.arc.nasa.gov>
Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
X-Mailing-List: <http-wg@cuckoo.hpl.hp.com> archive/latest/1036
> From:          hardie@merlot.arc.nasa.gov (Ted Hardie)
> Date:          Fri, 5 Jul 1996 12:49:40 -0700 (PDT)
> 	I think we have a terminology problem here.  HTTP/1.1 is not a
> new protocol.  It is a new standard.

Agreed.  I was too quick on the keyboard.

>  One of the main design goals for
> HTTP/1.1 has been to maintain interoperability with HTTP/1.0, which is
> an existing protocol, though not a standard.  The changes made to HTTP
> aim to maintain interoperability even at the cost of some design
> cruft; that's why there is a Host: header instead of a "simple" change
> to using full URLs in all requests.

That's because sending a full URL before knowing whether you're 
talking to a 1.1 server would break 1.0 servers.  Such is not the 
case for an origin server responding to a 1.1 client: it knows what 
kind of client it is dealing with.  Furthermore, even 1.0 clients are 
*supposed* to be able to deal with charset, which is allowed in 1.0, 
and almost all do.  Those that do not are not even 1.0 conformant, 
and have all but disappeared already.

Is it such a compatibility problem to require clients to at least 
parse charset (and drop it on the floor if that's the best they can 
do) before advertising themselves as 1.1?  I don't see that as any 
worse than requiring them to include Host:.

> We have to weigh that against the interoperability
> between 1.1 and 1.0, especially in situations where there are caches
> between the end-user client and the origin server.

Proxies already pass along charset if they get it.  Requiring (1.1) 
charset from origin servers instead of only allowing it (1.0) would 
not change anything in that respect.

> 	I was serious in my request that you write up your solution to
> this problem in an Internet Draft; we would all like this to work.  We
> are not trying to create something biased or broken.  

I don't think anyone is *trying*, but many are not paying attention 
to the existing brokenness, caused by the historical bias towards 

> We are trying to
> create something which will work well in a world in which not all
> clients, proxies, and servers are at the latest revision.  

A worthy goal.

> Sometimes
> that means we have to take things in several steps; 

The first step has already been taken - allowing charset in 1.0 - and 
has demonstrably failed.  The next step should be now, not pushed 
back to an indefinite future.

Francois Yergeau <yergeau@alis.com>
Alis Technologies Inc., Montreal
Tel : +1 (514) 747-2547
Fax : +1 (514) 747-2561
Received on Friday, 5 July 1996 14:44:00 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 14:40:17 UTC