Schizophrenia: RDF or MXF?

I have been interested in RDF developments and discussions for some time
now.
My "entrypoints" were digital libraries, classification, resource description, this interest is shifting towards knowledge representation.

I am getting the feeling that there is some schizophrenia in the concept of RDF.
The rest of this message is not the proven result of a well-thought alternative for RDF but merely a reflection of an uncomfortable feeling.
I hope the discussion on this item will yield a better scoping on the domain.

-----------------------------------------------------------

In the recent RDF documents I read the statement that "The goal of [the W3C metadata activity and] RDF specifically is to produce a language for the
exchange of machine-understandable descriptions of resources on the Web".

As the name of RDF strongly suggests a relation with resource description, I think the relationship with pure metadata description is a lot stronger.
Although resource description is an application area where metadata is heavily used, resource description is more specific than metadata description.
Not all resources are Web resources (most are not!), wouldn't it be more useful to have a description framework that is designed to describe all
resources?

As with all specifications a proper domain description is a primary requirement.
If this were given more precisely in the RDF specification documents it would have been obvious that especially the "resource" concept crosses a
number of conceptual boundaries, causing conceptual problems.
RDFS defines a resource as follows: "All things being described by RDF expressions are called resources."

I will suggest the following (coarse and incomplete) layered domain structure:

==========================================================
FRAME LEVEL: defines basic frame theory elements

FRAME    SLOT   FACET

==========================================================
META LEVEL: defines metamodel elements

NAMESPACE   CLASS  ATTRIBUTE   RELATION   INSTANCE   VALUE

==========================================================
ONTOLOGY LEVEL: defines theories, concepts and relations

THEORY    CONCEPT    RELATION

==========================================================
DESCRIPTION LEVEL: defines resource descriptors

==========================================================
IDENTITY LEVEL: defines virtual identities

==========================================================
PHYSICAL LEVEL: the resources themselves

==========================================================


The current RDF specifications suggest that the DESCRIPTION LEVEL is its domain while actually the META LEVEL is what RDF is all about.
The RESOURCE concept in the DESCRIPTION LEVEL (e.g. a book, a person) is very different form the RESOURCE concept in the META LEVEL (e.g. a class, a
property).

I would suggest renaming RDF in MXF (Metadata eXchange Framework).
The term "resource" will have to be reserved for the DESCRIPTION LEVEL.
This should avoid discussions on URI generation as identities for "real" resources are usually assigned by human interaction in contrast to nameless
RDF statements in database-dumps.

Regards,

Rob van Dort

Received on Monday, 1 May 2000 15:07:51 UTC