> -----Original Message-----
> From: ext Jon Hanna [mailto:[log in to unmask]]
> Sent: 16 January, 2003 15:23
> To: [log in to unmask]
> Subject: Re: rdfs:isDefinedBy revisited
> 
> 
> > I don't think it is well defined in any context. And I think
> > that the issue of what is a valid representation of a resource
> > or that one never can actually access the resource itself is
> > the very crux of the apparent friction between the Web and
> > the Semantic Web.
> 
> !
> Firstly I don't see what this has to do with the semantic web 
> (apart from it
> being part of the web).

Because most folks think that a URI denotes what they get
when they dereference it, and that is not necessarily what
it actually denotes to a reasoning engine. Yet if they 
start making statements about what they got, using the
name of "where" they got it, then they introduce noise
into the SW.

If I have a URI that denotes the city of Paris, and someone
dereferences it and gets a representation that is a photo
of Paris, and they say "Paris is out of focus" when they
really meant to say, "The photographic representation of Paris
I got is out of focus" then the SW becomes a repository for
garbage, not knowledge.

So clarifying what URIs denote and the distinction between
resources and representations and the URIs used to denote
resources versus representations is paramount to making the
SW work as a layer upon the Web.

> Secondly how can possibly you access anything other than a 
> representation?
> The concept of resource doesn't rule out the possibility that 
> a resource may
> be one and the same as what is accessed (a clear example 
> being downloaded
> software obtained from a uri designed purely to be were the 
> software lives).
> It just recognises that this isn't always the case. It's a 
> recognition of
> the way things are, not a statement about how things should be.
> 
> I most likely get a different response to deferencing
> "http://www.google.com/" than you. (Unless you are also in 
> Ireland). That
> doesn't change the fact that <http://www.google.com/> is resource.

For many use cases, content negotiation is a good thing. But what if
I need a precise copy, bit for bit, of a digital resource? Or if I
need to make a particular statement about a particular resource
(not representation) such as a checksum or digital signature?

The standard conneg reply is that you can ask for what you want, but
this is not anywhere nearly precise enough for many, many use 
cases since there is no standard saying exactly *how* you can
say what you want (MIME type doesn't cut it) and there is no
official concept of a canonical or platonic representation which
is *the* representation unless you say otherwise, and how a
server tells a client that it has returned (or not returned) such
a canonical representation.

For digital resources, that canonical representation would be a
bit-for-bit copy (not a transformation, not a summary, etc).

For non-digital or otherwise abstract resources, the owner of
the URI would be able to specify which is the canonical representation.

But this is, I agree, a completely separate issue that doesn't need
to live on this particular discussion list...

Cheers,

Patrick