Lee Iverson wrote:
> In message <3AE9DDF3.B4ABC642@eng.sun.com>, Eric Armstrong writes:
> >I think most of us agree that namespaces, as they have
> >turned out, are looking less and less like a real
> >solution, and more like a problem...
> Actually, namespaces were a real solution. The problem is that they
> bring the *real* problem into sharp relief. The need for some means
> to define, manipulate and match ontologies.
XML Namespaces (as opposed to 'namespaces') are what I believe Eric
is referring to, and the XML variety (which is something that functions
only at a lexical level unless we live in pretend or magic land) has
opened up at least as many ambiguities, mistaken assumptions, and
functional/operational nightmares as one could imagine. If there had
never been scoping, defaulting, and the confusion over attribute
namespaces, the tradeoff would have been more reasonable. We also
needed prior to their specification an XML packaging solution, such
that resolution of the namespace URI might have had some real
functionality. I agree we gain a measure of identification, but this
could have been accomplished in a different way. At least most people
I've talked to over the years think the Namespaces in XML Recommendation
was both under- and over-specified.
> Schemas and namespaces give us a way to define terms and locate their
> definitions unambiguously. They do not, however, provide any means of
> coordinating definitions and discovering commonalities in the
> ontologies that the names are referencing into. Schemas *are* useful
> but they stand pretty much at the level of programming language data
> structures. Saying that two applications are implemented in the same
> programming language has *never* been a guarantee of interoperability.
> I don't know about the Topic Map response to this need, but I know for
> a fact that sharp awareness of this fact is one of the driving forces
> behind the work on a service description language for DAML.
Topic maps provide the ability to unambiguously define something using
a URI just like XML namespaces, except this isn't mixed up in the
lexical level of specification, but is part of document content. We
essentially do everything RDF does in providing graph structures
associated whose nodes and arcs are associate-able using URIs, but
with the added semantics of topics, topic characteristics, scoping,
etc. that come with the ISO 13250 standard. I don't think of the topic
map syntax as really a metadata syntax in the same sense as RDF, but
it's certainly capable of associating metadata with an addressable node.
Though it's temporarily on hold due to some deadlines right now, I've
been working on a topic map representation of the Cycorp ontology,
and hope to eventually publish five or six topic maps, including the
taxonomy of Cyc. I've talked with Doug Lenat and I think we'll be able
to redesign some of the Cyc assertions that include variables so they
can be included in the XTM version.
Murray Altheim, SGML/XML Grease Monkey <mailto:firstname.lastname@example.org>
XML Technology Center
Sun Microsystems, 1601 Willow Rd., MS UMPK17-102, Menlo Park, CA 94025
america was once a paradise
of timberland and stream
but it is dying because of the greed
and money lust of a thousand little kings -- archy (1927)
This archive was generated by hypermail 2.0.0 : Tue Aug 21 2001 - 17:58:06 PDT