Re: [metaDataInURI-31] targets of sentences, was Re: Initial draft finding for public review/ comme nt.

Tim Berners-Lee wrote:
>
>
> On Monday, Jul 14, 2003, at 15:54 US/Eastern, Jonathan Borden wrote:
> [...]
> > This type of resource directly corresponds to the RDF "anonymous
> > resource"
> > or "b-node". Thus RDF b-nodes seem to be what Roy describes as the
> > target of
> > a sentence. I will go as far as to say that most if not all interesting
> > resources are anonymous.
>
> True that any relational database data which is mapped into RDF becomes,
> for example, a set of bnodes.   This is data, but it not using the
> power of the semantic web.  When you give things URIs, you
> can give them HTTP URIs, and those can be looked up. That is
> in fact useful.  This is the "web" bit of the semantic web.

I didn't mean to imply that resources which have URIs aren't interesting
(though I practically said that). I meant to reinforce the idea that targets
of sentences are resources just as are resources directly identified by a
URI. By "interesting resources" I mean  people, places and physical objects.
Such resources *might* be directly identified by a URI, but just as well
might be indirectly identified as the target of a sentence or a formula (as
you very well point out).

What makes such indirectly identified resources so interesting to me is that
they include anything that does not yet have a name as well as things that
have natural language names (but not *yet* URIs). The resources that we
already identify with URIs are already on the web -- by considering
resources that don't (yet) have URIs we have a much larger semantic web.

Concerning [metaDataInURI-31] I'm not conviced that there is any pressing
need to treat URIs as other than opaque (aside from the resolution process).
If we are going to assign meaning to URIs based upon part of their syntax
such as their scheme (which also falls into issue [httpRange-14]) I think
this would work only if we changed specs like XML Namespaces, to require
that namespace names have a trailing '#' if they use the "http:" scheme etc.
On the whole there'd be too much retrofitting of the current web, XML
applications, etc. to make it practical?

Jonathan

Received on Tuesday, 15 July 2003 20:26:11 UTC