Thanks for your comments, and I’m hoping you elaborate further.

My motivation for this particular “myth” was to somehow address what I feel to be a false and unfortunate tribalness between RDF and OWL. To me, all of this represents a spectrum from a relatively simple data model (RDF) through more expressive semantics until a relatively complete language emerges (OWL), indeed with many dialects.

Depending on purpose and use case there is no real conflict here, but actually a difference of purpose and use case that warrants explication. Yeah, perhaps it is too nuanced, but I think that better than false dichotomies.

You use many hot terms including DESTRUCTIVE, isn’t useful, danger, honestly. I welcome your elaborating on these concerns, because all I see now are assertions.

I also do not understand the statement that “OWL is not a graph. Period.” Yet we know that OWL (both 1 and 2 with explicit volumes for each) can be expressed as RDF. If I fundamentally misunderstand how OWL can be expressed as a graph I would like to know it.

IMO we are now seeing a transition from rather simplistic linked data with (sometimes) naive use of OWL evolve and require more sophistication and expressiveness. The entire attempt with OWL 2 and profiles, I think, is to help make more explicit a range of expressiveness of the language suited to different purposes and use cases.

I do welcome a dialog on these issues, because I think they are at the heart of current discussions and thinking.

IMO there has been a “tweaking” of the semantics of both OWL and RDFS to make them compatible; where RDFS is a layer of semantics applied over a data/knowledge representation framework (RDF) based on triples. To say that OWL is an extension of RDF is kinda like saying that the works of Shakespeare are an extension of the alphabet. I mean… yeah… formally that’s true, but it isn’t a useful thing to say, or a useful way to think about it.

The danger lies in the natural conclusion to this argument, which is that OWL is a graph. It isn’t. Period. So… I honestly think that it’s important to keep OWL separate from RDF, and simply say that RDF is a way of representing OWL, in exactly the same way as RDF is a way of representing my stock portfolio.

Just my opinion… but it’s an opinion based on several years of trying to help students in my lab understand what OWL and RDF are. Keep things as simple as possible, but no simpler

– Organic classification
– Statistical relevance
– Social rankings
…

I have the feeling that RDF, OWL, and SPARQL are just too “academic” and the history of Internet has shown us that the “community” (developers and users) prefers simple non-academic solutions. For example… http://www.delicious.com... offer a clear benefit for the user (save your bookmarks online) and out of that simple egoist benefit TONS of emergent knowledge can be extracted:

– Classification of sites (actually a machine can now understand what a site is about based on the statistic probability of its tags)
– Popular links
– Popular tags
– Treds
– etc etc

I still remember how OpenGIS was the “academic standard” that was going to be the future of online mapping applications… and the BOOM! Google Maps came out and transformed forever the world of online mapping.

So… let’s not forget that all these semantic web standards come from academic origins and not from the natural evolution of the Web. The future will be much simpler… I assure you

An interesting link:
http://lexandera.com/2009/04/crowdsourcing-the-semantic-web/