Re: Information resources

mån 2007-11-26 klockan 10:48 +0000 skrev Dan Brickley:
> One of the ways we can be clearer about RDF/OWL class definitions, is by 
> being clear about which things are *not* in some class we define (and 
> ideally also giving some explanation). 

Yes, I agree.

> For example, considering "books", 
> which of the layers of the FRBR conceptualisation (work, expression, 
> manifestation, copy) are information resources in the AWWW sense, and 
> which aren't.

Perfect example, thank you.
> 
> RDF terms are also slippery things, but at least they're a W3C creation, 
> so getting an account of whether RDF terms are "information resources" 
> should be a smaller task. I'm yet to hear a knock-down argument as to 
> why RDF terms (dc:creator, foaf:OnlineAccount etc) are not human works 
> with http-accessible representations, just like Web pages and books. I 
> don't really mind putting in the 303 redirects, but I still haven't come 
> to any clear rule-based or intuition-based sense of what things are in 
> or out of the awww:InformationResource class. A lot of the http-range 
> discussion has been in terms of people, cars, etc; RDF properties are 
> much closer to the borderline, and so might be a useful category of 
> thing to help make these definitions sharper.

I happen to think Properties are non-IR, and I think there are
compelling arguments for that.

I think the problem is that people, cars, etc all have an existence
outside the Web - they are physical things. The class of things you are
thinking about seem to be abstract entities of different kinds, correct?

If you were to say that a Property is an IR, then that same argument
applies to all abstract entities. Concepts, classes, mathematical
functions, etc etc.

But even so - timbl's distinction between the content of a thing and
descriptions of that thing apply. That is:  a description is not the
same as the content of the thing. Or put in other words, what *would* be
the content of dc:creator? I can't find any message that would convey
anything other than the definition. 

One potential representation would be all pairs of (subject, object) to
which the property applies. Now, RDF Semantics defines a Property to be
something else than this set, so that is not appropriate either. So if
we can't imagine a representation of the *content*, it's not an IR.

/Mikael


> 
> cheers,
> 
> Dan
> 
> 
-- 
<mikael@nilsson.name>

Plus ça change, plus c'est la même chose