Re: URIs: resources and contradictions was: Re: httpRange proposed text

Jonathan Borden wrote,
> I don't claim that removing ambiguity about the mapping of URI to
> resource somehow removes all ambiguity in the system ... that would
> be a miracle. I think we need to move step by step and clarifying the
> relationship between URI resource and representation would be a great
> first step.

I agree that that's a valuable goal. But I'm not convinced that making 
the URI to resource mapping 1:1 actually promotes it.

Take http://www.w3.org/ again. You say that there's a 1:1 mapping of 
that URI onto a resource. You also seem to be saying that there are 
contexts where that resource "means" a particular document, and that 
there are contexts where that resource "means" the W3C's web site. 
Right?

I know what a document is, and I know what a web site is, but I've 
really no idea what a resource which might "mean" one or the other is, 
unless it's just an artefact of a semantic theory. Or try it again with 
another example: I know what a document is, and I know what a car is, 
but I've no idea what a resource which might "mean" one or the other 
is, again, unless it's just an artefact of a semantic theory.

I have no problem in principle with semantic theories helping themselves 
to such artefacts so long as they do useful work. But it really isn't 
at all clear to me what good they're doing here. As far as I can see 
all we have is an implicit two-level model, where all the interesting 
stuff has been shunted off to the second level and ruled as out of 
scope for RDF.

But now I'm beginning to lose my grip on the point of the exercise. If 
all the RDF MT is doing is regimenting the first, uninteresting, level 
(the one that isn't about cars or documents) then what's it for? If you 
can't saying anything directly in RDF about either cars or documents, 
then who cares about interoperable syntax? Why not just use KIF or 
first-order predicate calculus and be done with it?

When you say you want to see contexts in RDF, do you mean you want 
something like a possible worlds semantics, where each context is a 
world, the MT is extended to accomodate interpretations over sets of 
worlds, and the truth values of non-modal assertions are fixed relative 
to worlds?

That's a very attractive proposition, but it's such a radical departure 
from RDF as it actually is that I really can't understand what you're 
currently arguing about wrt resources. Changing the language and model 
in that way would change the nature of resources in RDF so 
fundamentally that we really wouldn't be talking about the same thing 
any more. Denying that would be a bit like saying that the integers 
sort of include fractions, because we can embed the integers in the 
reals.

Cheers,


Miles

Received on Sunday, 4 August 2002 06:33:07 UTC