Does every URI identify a resource? (Was:Re: namespace usage as assertions)

Dan Connolly wrote:

> "Manola,Frank A." wrote:
> >
> > This is probably a really dumb question at this point in the discussion,
> > but as it relates to the RDF model of assertions being attributable to
> > individuals, I'd like to get the point nailed down.
> >
> > To wit:  isn't it the case at the moment that, since the URIs that
> > identify namespaces may actually point to nothing,
>
> That's not the case. Every URI, by definition, identifies/points to a
> resource.
> URI means "Uniform Resource Identifier"; URIs identify resources.
> cf RFC2396 for the exact definition.

What RFC2396 says (Sec. 1) is: "Uniform Resource Identifiers (URI) provide a
simple and extensible means for identifying a resource."  To paraphrase: if
you want to identify a resource, a URI is the right tool for the task.  A
resource, according to RFC2396, is anything that has identity.  Examples are
electronic documents, weather report services, human beings, corporations,
and books in a library.

Dan, you're making the inference that "a URI can identify a resource" implies
"a URI necessarily identifies a resource".   I don't see the logical
justification for that inference.   A URI is a syntactic form, but nothing in
RFC2396, the defining document for URIs, says that it is necessarily anything
more than a syntactic form.   Usually when I write a syntactically correct
English sentence I intend it to convey some meaning, but there are
syntactically correct English sentences that have no meaning at all.

URIs may have identification of resources as their intended purpose, but that
doesn't prevent their being used for other purposes, any more than the
intended use of a shovel as a device for moving earth prevents me from using
a shovel as a walking stick.    The authors of the namespace spec appear to
have such an alternate use in mind, namely, URIs as strings that are
suggestive and can easily be made unique.

You might take the view, I suppose, that the very act of writing down a URI
creates a resource, even if that resource is entirely ethereal in nature and
has no meaning or definition beyond the URI that created it.   Nothing in
RFC2396 contradicts that view.  But you seem to be looking for more: that the
resource identified by a namespace name somehow provides information about
the namespace beyond what can be inferred from the namespace name itself.

According to the namespace spec there's nothing wrong with using
"mailto://connolly@w3.org" as a namespace name.  Presumably the resource thus
identified is your mailbox.  But how could that resource possibly provide
useful information about the specific names used in that namespace?  Is
someone supposed to send you email to ask for it?

Is there a real difference between a URI that identifies nothing useful and
one that identifies nothing at all?

Paul Abrahams

Received on Wednesday, 21 June 2000 17:50:25 UTC