Thinking about RDF and Topic Maps

Danny Ayers, in this blog entry, talks about issues related to representations in RDF that speak to issues I have thought about for a while now. I think that now is a good time to start a dialog between the RDF tribe and the Topic Maps tribe. It's a double-edged knife, one that cuts both ways, looking at the true nature of the inquiries of each tribe.

I like to think about it this way: the core of the topic maps inquiry is to
satisfy a couple of important use cases: finding and reminding. In
those two use cases lie two primitive notions: subject identity and
names for things. Those are the two primitives that topic maps place
front and center, whereas, it seems to me, OWL emphasizes inferencing in
subsumption hierarchies, relegating subject identity to "proper use of URIs". I like to think about subject identity in the same terms a lawyer might do so in a court case. There, properties of the subject, more so than some URI, become all important. A trial might turn on something as trivial as shoes worn on some particular day. As topic maps are evolving, particularly in the case of the TMRM (topic maps reference model), we are seeing more emphasis placed on comparable subject properties than on precise URIs, which, in many cases, do not (yet) exist. We are seeing the evolution of the ability to "confer" identity on a subject according to circumstances. I think this line of inquiry can map directly into rdf work.

Topic maps (indeed, "subject maps") add one important consideration
outside subject identity and names for things: a guarantee that any
proxy for any subject (aka Topic), is the one place you need to go (in
*this* map) to find all that is knowable about that subject.

The knife cutting in the other direction suggests that, at the implementation level, topic maps could evolve along lines suggested from rdf work. Indeed, some of my own work involves the use of Jena coupled with JDBM for a backside.