RE: Asunto: Re: [www-rdf-interest] <none>

Hamish wrote:
> I promised myself I'd lurk a while, but as ever I can't resist. Hole
> picking welcomed, I'm still on the steep part of the learning 
> curve and
> need every bit of help I can get.

Hi Hamish, I thought I saw you lurking there ;-)

> ... but I'd hesitate to use "anti-social" in this context.

I think what we are all talking about is good / bad practice rather than
something that warrants the use of 'anti-social' and 'burden on the
community'.

The notion of subclassing and declaring equivalence to link terms in
different schemas is so baked into RDF that to say 'this is a great
feature of RDF that will be important for implementing semantic web
applications' and then 'by the way, don't use it because it's bad
practice' seems odd. As most non trivial schemas will use these features
and RDF tools and software will have to have the ability to deduce
semantic equivalence, I take the view that it doesn't matter whether we
re-use or invent (with appropriate linkage) and it boils down to
personal style and preference. When tackling a particular piece of RDF
data, whether the RDF software uses it's mechanisms and algorithms for
semantic linkage is here nor there - it still has to have that ability.
And whether the data re-uses or uses equivalent terms to 'known' terms,
the result of the processing will (hopefully) be exactly the same. As I
mentioned before, RDF software (that works) doesn't care either way or
care about what we humans think is good or bad practice.

Let me describe a scenario. A domain expert designs an RDF schema and
uses rdfs:subClassOf and rdfs:subPropertyOf for various terms
extensively and legitimately because he wishes to refine existing terms
to more accurately model his domain. Then he comes to 'date'. Should he
use dc:date and introduce just one term from the Core into his vocab or
subclass dc:date and preserve a consistent namespace across his schema.
What would you do? I personally would subclass. Others may re-use.
Either way is perfectly valid and ultimately it really doesn't matter.

Received on Wednesday, 14 January 2004 08:15:59 UTC