W3C home > Mailing lists > Public > www-tag@w3.org > July 2002

RE: httpRange-14 , what's the problem

From: Bill de hÓra <dehora@eircom.net>
Date: Thu, 18 Jul 2002 09:59:33 +0100
To: "'www-tag'" <www-tag@w3.org>
Message-ID: <000701c22e39$6f8f96d0$1fc8c8c8@mitchum>

> -----Original Message-----
> From: www-tag-request@w3.org [mailto:www-tag-request@w3.org] 
> On Behalf Of Paul Prescod
> Sent: 18 July 2002 02:36
> To: Joshua Allen; www-tag
> Cc: Tim Bray
> Subject: Re: httpRange-14 , what's the problem
> I know you were trying to help but I don't think you 
> clarified the argument much!
> Joshua Allen wrote:
> > 
> >  The
> > proponents of expanding the range of http are making three 
> generalized
> > arguments:
> As Roy pointed out after your message, the range of HTTP is 
> already far more than documents.
> > A. Some people claim that *all* resources which one would care to 
> > identify can (and should) be dealt with through REST, and therefore 
> > rule #2 applies.
> This is not an argument for expanding the range of HTTP. It 
> is an argument for expanding the *usage* of HTTP.

Can we please make a distinction between the HTT Protocol and the http:
URI scheme? Conflating the two isn't helping matters.

RDF knows nothing about HTTP, or for that matter, the Web. That's not
debatable. Strictly speaking RDF knows nothing about http: either,
though you can use the http: scheme on its own to name all RDF resources
to the exclusion of all other schemes (cue café philosophy about
countably infinite sets). 

RDF's sole requirement in naming is things are uniquely named. Using
URIs is a convenient way to do that, but you could take URIs out of RDF,
replace them with another naming system and the RDF MT would still make
sense, so long as you meet the uniqueness criteria. There's no question
that URIs are the best way we have to name things, but they're not
necessary. URIs are a naming policy for RDF.

When RDF is reasoning with URIs it is reasoning about resources, not
HTTP/REST representations. With RDF it doesn't matter what URI
scheme/policy/religion you want to subscribe to name your car, RDF is
talking about your car, not about your HTTP representation of a car.
That's not simply a pedant's outlook. As Roy says there are no resources
moving through the Web, just representations. Web machinery is in the
business of processing and distributing representations. RDF machinery
will not be working at that level. 

Where RDF is a bust is when you want to use the same URI to name many
resources. The reason that one to many mapping's  a bust is simple: you
can't merge two RDF graphs and assume the entailments aren't bogus
unless you know all the things named in the two graphs are named
uniquely. Graph merging is one of the most useful things we will want to
do with RDF. RDF has no processing model that we could use to
disambiguate matters. 

Bill de hÓra


Received on Thursday, 18 July 2002 05:00:32 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:55:52 UTC