« Getting closer to releasing Wilbur2 | Main | RSS in Longhorn »


Two Towers

Some very disturbing developments have emerged in the Semantic Web community. Among the ongoing discussion on rules are proposals that will split the Semantic Web, as far as semantics goes. Newly seen versions of the "layer cake" suggest new approaches to how we are going to structure the technologies for the Semantic Web. This one (from Tim Berners-Lee's recent keynote talk) is particularly interesting:

New Semantic Web Stack

A recent paper from Horrocks et al is an excellent review of the current situation. Given that many existing rule languages (and proposals for rule languages) have Datalog semantics, we end up with the following situation (per my reading & understanding of things):

  1. Datalog cannot be layered on DLP, unless you adopt different semantics for DLP (for the purposes of this story, let's call that DLP-Datalog)

  2. DLP-Datalog cannot be layered on RDFS

  3. OWL cannot be layered on DLP-Datalog

  4. DLP-Datalog is not compatible with RDF semantics

From all this we have to conclude that it makes more sense to layer all this new stuff on top of XML (or some other syntax) and forget about RDF and OWL (at least as far as their semantics go - funny, I always felt semantics was the important part; Semantic Web and all that, you know). Does this all sound scary? You betcha!

In a way, the Semantic Web community is in a danger of splitting into a camp that makes the Closed World Assumption (and the Unique Name Assumption) and to another one that doesn't. So far I haven't seen any unifying "logical framework" that some people have been suggesting. And despite the fact that we originally started building RDF with, say, OO models and databases in mind, I firmly believe that the Web needs the "Open World" Assumption (even if building systems that make that assumption may be harder).

Of course, apart from the Semantic Web (or KR community), the other significant stakeholder of possible future W3C work on rules is the "business rules" community. There, rules are more like a programming paradigm rather than KR, and my take on this is that it may be foolish (or at least quite optimistic) to assume that both communities could be satisfied with the future outcome of such an activity. I may be wrong, but I am assuming that in the business rules community there's nothing wrong with the Closed World Assumption (in fact when dealing with databases it is sort of the natural and desirable way to go).

All this makes me worried. Fortunately, the recent W3C workshop on rules did note in their report that

"The workshop gave many indications that a W3C Recommendation here would be useful, but it was less clear what sort of standard would satisfy a sufficient base of users. In any Activity Proposal following from this workshop, a Working Group should be given a clear and narrow scope, making it easy to determine its relevance to various parts of the greater rules community."

But still, I fear that not all people see the gravity of the emerging situation. My "conspiracy theory" is that (given the long-standing controversy surrounding RDF etc.) the opponents of the Semantic Web will now be able to claim that "see, you could do it with just XML", however stupid such a statement actually is.

And besides, I never liked Tolkien anyway...

Posted by ora at 06:54


broken link to horrocks, thsnks

Posted by: lkjlkjlkj at June 11, 2005 10:35 AM

Ora, that Horrocks link is ""... do you have a link/citation? It's been mentioned on IRC.

Posted by: Rich at June 11, 2005 12:09 PM

Broken link fixed. Sorry.

Posted by: Ora Lassila at June 11, 2005 01:47 PM