Updated: corrected misattribution from Shelley Powers to Dorothea Salo.
Many times I have argued with Mike Champion, in private and in public, Semantic Web technologies are hardly some crazed figment of academic imagination. I don’t mean the grand Semantic Web dream of intelligent agents weaving an autonomous Web with minimal human meddling. I’ve always said that Semanitc Web technologies are very useful in immediate, less ambitious context. They are indeed the next common-sense step from XML syntax. One of my core arguments has always been that the success of XML drives directly towards SemWeb whether or not the journey goes through significant bodies of explicit RDF data. I was surprised to learn that last week Mike Champion wrote:
FWIW, my recent epiphany that the Semantic Web  stuff is not as
wacky as I once thought came partly through the possibly heretical
thought that there is no need to convert much actual data, and perhaps
not much metadata, into RDF. The powerful Semantic Web idea for me is
the notion of an ontology — meta-metadata if you will — that relates
concepts and relationships in real-world metadata (service
definitions, XML schemas, data dictionaries, etc.) to one another in a
way that supports automated inference. DAML-OIL and OWL are RDF under
the covers, I guess, but that doesn’t strike me as any more
interesting than the fact that they are serialized as XML. The power
AND the challenge comes from the modeling exercise of defining a
network of resource-predicate-value assertions that can be navigated
to automate useful tasks.
I suspect that “epiphany” was just a vague re-remembering of the long-standing arguments of those of us, the XML people who have never had much of a problem with RDF (excepting the horrid XML syntax). Not just me, but also Eric van der Vlist, Bill de hÓra, Joshua Allen, and more recently Norm Walsh, Roger Costello and Bob DuCharme. Someone (I think Bill de hÓra) used this group as a counter-example to Dorothea Salo’s infamous rant against RDF absolutism.
Personally, I have never looked at RDF without an eye to making explicit RDF unnecessary given well-designed XML. It has been almost four years since we designed the XML <-> RDF mappings in 4Suite. It’s not perfect, but the idea is that data is stored in XML and automatically mapped into a back-end RDF store as basically a metadata index that XML people need not worry about after the initial schema design. I’ve had a lot of discussions with folks like Paul Prescod and van der Vlist over the years about designing some sort of RNG schema annotation for this purpose, and I’ve seen parallel efforts along such lines from others.
I’ve tended to argue that a mapping to RDF may be the easiest way to SemWeb, given all the very good work that has been put into RDF, but I’ve never let my imagination be limited by RDF in its best-known forms. In my Thinking XML column on IBM developerWorks I have a long-running (almost 30 article now) development of the ideas of XML semantic transparency, and the fact that all the trappings of the Semantic Web become low-hanging fruit once one achieves semantic transparency (more easily said than done), whether or not RDF is the means to this achievement. OASIS core components or XML Topic Maps could be as much a path to SemWeb as RDF, although in my experience, RDF is simpler and much more “hackable” than either of these (or any other alternative I’ve seen).
I’ve hear a lot of castigation of Semantic Web technologies over the years. Of the many complaints I’ve heard, Mike Champion’s have always been the hardest to credit. He would argue from the point of ostensible “practicality”, along lines that would never have admitted the success of XML 1.0, and which were always contradicted by the growth and energy of the RDF community. This made it hard to understand where his “practice” was coming from, or what universe of developers it intersected (certainly not much of the very diverse universe of developers I’ve been lucky enough to work in).
One funny thing about the timing of Mike Champion’s posting is that it was two years ago that I first saw the idea of Semantic Web from XML in a more mainstream context. It was a presentation by William Ruh of Software AG, where Champion worked at the time. Ruh pointed out (and I was happy to hear a Veep type say what XML veterans had been saying in narrower XML circles) that the many industry efforts to standardize XML semantics were a path to the Semantic Web. I found other interesting people at Software AG (including Klaus Fittges, CTO) who saw practicality in and were looking to profit from SemWeb technologies.
In the end I keep coming back to my mantra. RDF is not an XML syntax. RDF is not a DBMS system. RDF is not an AI or inferencing system. RDF is just a specification of directed, labeled graphs (circa Somp Sci 201) with URIs as edge labels and URIs or literals as node labels. That’s all. I would claim that almost all abuse of RDF has stemmed from failure to understand this point. Maybe it’s too subtle a point? Seems clear enough to me, but I maybe I’m missing some complicating perspective. In any case it is nice to see others begin to come around to it (a lot of people have been taking off Semantic blinders lately, I’ve noticed). Welcome to the club, Mike.