RE: URIs as names (was: What do the ontologists want)

Larry Masinter wrote:

>
> reductio ad absurdum doesn't really help here. Personally,
> I think the fuss is just because people haven't figured out
> how the mystical power of the URL turned improverished hypertext
> into the multi-billion-dollar World Wide Web.
>
> The use of URIs in RDF and as XML namespace names to identify
> things other than network resources is a bit of semantic extension
> that doesn't work all that well.

When you, as a coauthor of RFC 2396, wrote that a resource may be anything
that has an identity, it didn't limit the use of any particular form of URI
(e.g. http scheme based) to identify only particular types of resources
(e.g. network resources).

For XML namespaces, I think its a perfectly good fit because we can place a
network retrievable document at the end of the namespace URI that describes
the namespace (http://www.rddl.org). When a resultant entity is identified
to be a RDDL document, the exact intention is that the entity is a
description of the namespace.

> You can certainly imagine
> defining a URI scheme that would work to name a grain of
> sand on a Pensacola beach, but is it really a good idea to
> do so? It's kind of like using Goedel numbers instead of
> equations.

the problem as i see it, is that we need to better define the intended
relationship between URI, resource and the set of entities that result from
resolving the URI.[1]

> RDF supplies yet another context, but currently isn't clear about
> what it intends the meaning of the URIs within it to be.
>
> Part of the problem is that there's a lot of sloppiness about the
> level of quoting or indirection. Some RDF statements seem to be
> about the actual resources that a Uniform Resource Identifier
> identifies, and others are about resources that are merely
> described or correlated with those other resources. Getting
> the level of indirection right is hard; we haven't been explicit
> about when EVAL is called.
>

yes, exactly. another problem is the relationship of URIs -> Resources
(intended to be 1:1) yet there is no mechanism to directly relate Resources
<-> Resources. The need to do so arises under the following (very common
circumstance).

suppose the URIs (some ficticious)

1) http://www.w3.org/People/Connolly

2) http://dm93.org/

3) http://hcfa.gov/isRegistered?ssn=111223344

in 1), the W3C asserts that this URI describes "Dan Connolly"
in 2), someone named "Dan Connolly" who lives at ... (i.e. per WHOIS)
asserts that this describes the family of "Dan and Mary"
in 3) the US government healthcare authority asserts whether a person with
the SSN=... is 'registered' (this is ficticous)

So, we have 3 proper Resources

1) [a Person such that [worksAt w3c.org, name "Dan Connolly]]
2) [a Person such that [name "Dan Connolly", [marriedTo [a person such that
[firstName "Mary"]]], [livesIn "Texas"]]]
3) [a Person such that [SSN "....."]]

What we need is a way to _assert_ that these resources each reference the
_same_ resource (the actual person with an identity).

Topic Maps and the TM Processing Model might provide such a framework. Under
the TM paradigm, 1) - 3) would refer to a "topic" and "the actual person"
would refer to a "subject".

Performing the same task in RDF would entail encoding the statements 1) - 3)
as URIs and perhaps using something akin to daml:equivalentTo ... but for
instances... to relate the intention that the resources somehow refer to the
same resource.

-Jonathan

Received on Tuesday, 22 May 2001 08:29:44 UTC