W3C home > Mailing lists > Public > www-tag@w3.org > January 2003

Re: yet another sidetrack on what a URI identifies

From: Roy T. Fielding <fielding@apache.org>
Date: Wed, 15 Jan 2003 15:27:19 -0800
Cc: <sandro@w3.org>, <www-tag@w3.org>
To: Patrick.Stickler@nokia.com
Message-Id: <E430C060-28E0-11D7-92D1-000393753936@apache.org>

> That may be true for some folks, but some of us simply had a problem
> with ambiguity of xmlns URIs being presumed to denote schemas, models,
> vocabularies, and other resources which are far more than "sets of
> names" and if a namespace URI denotes that namespace, then I would not
> consider an XML Schema, or RDF Schema, or RDDL document as a valid
> representation of that namespace, since they all invariably include
> a great deal more than embodied by the namespace itself, which is,
> after all just a simple set of names that infers no semantics, 
> structure,
> or expectation of usage on those names.

Almost all representations contain more than the subject of the 
representation.
The representation is not supposed to be equivalent to the resource.  
It is
supposed to describe the state of the resource.  As such, the 
description
often includes references to things outside the resource itself, just as
a description of me would invariably mention California.

> Dereferencing a namespace URI and getting a RDDL document or XML Schema
> is like having a URI that denotes Paris, and getting a "representation"
> that also includes all the information about Europe as well.

Most pictures that attempt to describe Paris will, in fact, include the
surrounding countryside.  The fact that the representation isn't 100%
true to the reality of Paris is inevitable -- that is the nature of a
network-based information system.  Even a resource that can be described
as a virtual document has a reality that is separate from the individual
representations that are delivered upon request.  That is why RDF can
use any URI to define a subject, and why HTTP can transfer a
representation of anything that is given a URI -- because what HTTP
transfers is not the resource, but information describing the state
of the resource.  What RDF cannot do is use the URI alone to make
assertions about the representations and assertions about the resource.
It must distinguish the two because they are different things, even
if they are both identifiable (directly or indirectly) by one URI.

If I ask for today's weather in Los Angeles, is it really so bad that
I get more than just today's weather in response?

What the author is trying to identify is hopefully embodied in the URI.
How the server responds to a GET on that URI is variable, but hopefully
consistent in including the state of the resource that the author 
intended.
There is a potential for error on both sides, which can be reduced by
further clarification in human-readable text or something like RDF in
regards to both the authority's intended subject of the URI and the
persistent characteristics of its representations.

....Roy
Received on Wednesday, 15 January 2003 18:27:43 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:47:15 GMT