RE: a theory of meaning for URIs

> > In this theory, any other meaning or interpretation
> > comes from the context. If some representation
> > system wants to add some additional meaning to
> > the URI, that additional meaning comes from the
> > representation system, and not the URI itself.
> 
> Well, that makes this theory pretty uninteresting;
> the question we've been asked to discuss is how
> representation systems, particularly RDF, contribute
> to the meaning of URIs.

Well, it relieves you from having to decide
on any intrinsic semantics of the URI beyond
the semantics already defined. If RDF
is adding meaning by using URIs, then RDF can
define that meaning, without having to discover
and define any intrinsic meaning.

RDF could say "any place in RDF where I want to
talk about a concept, I can use a URI which
should point to a web page which describes the
concept". (This is the 'implicit tdb' theory).

In this formulation, RDF assertions about W3C
(as a consortium) could just use 'http://www.w3.org'.

RDF assertions that wanted to talk about the
web page at http://www.w3.org would have to
use something other syntax. I think 
    data:text/url,;http://www.w3.org
is pretty ugly, but perhaps you could either
come up with some other syntax, or else define
rdf:about to be an exception, or some other
formulation.

However, I think that no theory of meaning
of URIs in RDF should be allowed to avoid
being clear about whether a reference to
http://www.w3.org/People/Connolly/ is about
you or about your web page.

> > However, I would prefer updating the definition of
> > the 'urn' scheme to make it clear that, for a URI
> > that starts with "urn:", that the next thing in the
> > URN syntax is a token identifying a 'namespace
> > authority'.
> 
> I don't see any motivation for treating the
> bit after urn: as different from the bit
> after http:// , nor for treating that differently
> from the xyz bit in http://example/xyz/
> and so on.

Because the definition of 'http' is tied up with
the HTTP protocol, while the definition of 'urn'
describes the role of naming authorities. You treat
them differently because their definitions are
different.

I'm not sure I can 'motivate' this in isolation.
My main motivation is that you don't have to provide
a stronger theory of meaning for URI than the one
that is already defined. 

Any theory of meaning that relies on all URIs having
an 'owner' and requiring the owner to 'say' what the
URI means is weak. I'm trying to suggest an alternative.


Larry

Received on Thursday, 25 September 2003 17:10:58 UTC