RE: Clarifying what a URL identifies (Four Uses of a URL)

> We agree that with HTTP a number of different
> representations of the thing identified by the URI.
> I want to use the URI to identify the picture.
> Roy has always felt it identifies the car.
> Either system is self-consistent.

This is, for me at least, the crux of the problem. If
I have a URI that denotes some resource and everyone
agrees on the denotation of the URI, and that resource
is in fact a digital resource (e.g. a specific RDF Schema
instance) and my SW agent needs to interact with that 
resource (obtain and evaluate the statements expressed
in that schema) it seems fully valid according to the
specifications that an HTTP server could return as
a representation of that schema a JPEG image depicting
the pages as printed on my 70's vintage line printer.

I.e., there is no concept that I am aware of for the
canonical representation of a digital resource (i.e.
a bit-for-bit exact copy) such that one can both name
a digital resource and obtain a reliable copy of that

There is also insufficient discussion of how far a
given representation can be a superset of information
not embodied in the resource denoted by a URI.

I am uncomfortable with RDDL instances being returned
as representations of XML namespaces because (a) they
do not in fact (necessarily or usually) enumerate the
terms grounded in that namespace (and I consider such
an enumeration a necessary component of any complete
representation of an XML namespace) yet may further contain
any arbitrary information about any other resource even
remotely related to that XML namespace or some term
grounded in that namespace. It seems to me that every
example RDDL instance I've ever seen is not a valid
representation of the XML namespace -- as I would 
perceive a valid representation.

For me, here is where the interface of the Web and
the SW breaks down: the lack of a mechanism for 
obtaining canonical representations of digital
resources and a more precise definition of completeness,
economy, and precision as those properties relate to
representations of resources. At least TimBL touches
on that last point in his action item to the TAG

by saying

   ... an HTTP URI ... must be used to refer to a unique 
   conceptual object whose various representations have 
   a very large a mount in common ...


   ... the difference between different representations of 
   the [resource] is very small ...

Though it seems that many folks do not have such a 
constrained view of how broadly variant representations may
diverge from one another or from the inherent properties
of the resource denoted.

I think that something far more comprehensive, and
most importantly, authoritative should be provided
by the TAG on both of the points identified above.

It seems to me that a reasonable and useful guideline,
though somewhat imprecise, would be that the intersection
of commonality between representations should be far
greater than their deviations and that the common intersection
of all representations should completely or very closely
coincide with the inherent properties of the resource
denoted. And for digital resources, there would be at 
least one canonical representation which would be bit-for-bit
identitical to the digital resource itself.



Patrick Stickler, Nokia/Finland, (+358 40) 801 9690,

Received on Thursday, 23 January 2003 04:23:42 UTC