RE: web proper names

> So, if we try to resolve the meaning of URIs within the 
> framework of RDF, 
> perhaps we are just looking in the wrong place.

I agree with that. URIs mean whatever they mean first, HTTP, RDF and
other technologies act on them after that.

It's still on-topic here though because it affects RDF. In particular if
RDF is not compatible with the meaning of URIs then it is broken to its
core. Less drastically it has several implications for best-practice.

Really this debate comes from HTTP more than from RDF. But since RDF
gives us a means to talk about resources it tends to quickly end-up
being discussed using either RDF or pseudo-RDF. (For the same reason
that any sufficiently complicated web technology concerned with meaning
and/or intent soon becomes an incomplete and buggy version of RDF
anyway).

> So, what's needed?  I'd argue that our discussions about web 
> proper names 
> and URIs can never converge to something computationally 
> useful, unless we 
> broaden the frame of discussion to bring in natural language 
> processing.  In particular, a solid way of computationally 
> linking RDF and 
> English, in both directions, would do the trick.

I disagree. One of the benefits of using URIs is that they can have
precision that isn't easily available in NLP even when that NLP is
performed by the human brain (which seems to do an okay job in this
regard most of the time). Working out what "set" means is a lot more
complicated than working out what ISBN:0321185781 means. URIs are closer
to the latter than the former.

This isn't to say that NLP <-> RDF isn't a very useful area of research,
but just that there's plenty we can do with RDF on its own, and we
should be permitted to do so.

Received on Tuesday, 21 September 2004 10:15:10 UTC