Thinking Clearly .

December has be an active time on the RIF working group mailing list and the OWL Dev list, especially about “conformance” (and compatibility). Conformance clauses of standards are one of the stickier tar pits since they serve (or are thought to serve) as the basis of interoperability. Put your TRUST in the MUST, as it were.

Conformance clauses are tricky for several reasons both technical and political. Technically, no standard is a tight enough spec to exactly determine everything about a system, nor would you want it too. Performance, for example, is typically an axis of differentiation of (conforming) products. Politically, every vendor (commercial or non) wants it to be as easy as possible for their system to fully conform. (Some may want it to be difficult for competitor systems to fully conform…but not always.) If the spec is ambiguous, it may be possible for two conforming systems to fail to be interoperable. Similarly, if the spec has optional aspects. Then there is the issue of determining or certifying conformance (who does it; what the procedure is; etc.).

Conformance clauses can be a handy thing for customers. For example, though obviously we knew how to implement Qualified Number Restrictions in Pellet—- and there was clear user demand for them—- we didn’t implement them until the pressure of OWL 1.1 came along. After all, it was work, not interesting or research supporting work, and in the end we’d have a implementation of a very boring extension (from the publication perspective) to the OWL language. Not exactly a strong motivator in the absence of specific funding. However, “fully supporting OWL 1.1” is nice and punchy and valuable to us.

I leave it to the interested reader to sort through the conformance debate on the RIF list. I made a little contribution which isn’t a bad place to start (though it’s not at the beginning). Michael Kifer defined a (contentious) notion of conformance for RIF systems and, in a reply to me, distinguished between conforming to a RIF dialect (which really is a property of documents) and implementing a dialect. (This just goes to show the trickiness…these raucous threads may have stemmed from a terminology confusion!)

The schedule for OWL Experiences and Directions is now up, and it looks good. Some of the papers are really very interesting, and all of them I think will contribute to good discussion.

Clark and Parsia (us!) are sponsoring it, though we’ve not given our new logo for the OWL ED website yet. I blame Kendall!

The most important session, in my mind, is toward the end of the first day, “Ongoing and Future Standardization Efforts”. Here is where we lay the groundwork of what we are going to do over the next year. For me, I hope in the next year we can reach broad, implemented consensus on SPARQL/DL and DL Safeish rules. Oh yeah, maybe get a new round of standardization going.

OWL 1.1 is proceeding apace. FaCT++ and Protege-OWL support it. Pellet and (prelim) Swoop support should be ready by OWL ED (go Evren go!). I’ve not heard from the Racer guys, but they had a bit of a head start.

In a very minorly unusual move, OWLED is rather actively soliciting demos and posters right up to the demo/poster session. In fact, if you get a one page description of your demo/poster in by Oct 20th, you still have a shot at the proceedings:

The workshop will have a session for software demos and “posters” (formal or informal) describing work in progress. All participants, even those without a paper in the program, are invited and encouraged to demo their tools and/or present a poster. Interested parties may send the organizers a one page description of the demo/poster before October 20th. Such a description is not required, but descriptions submitted prior to October 20th will be considered for inclusion in the proceedings.

So, come out!

There are a number of “rules” systems, where what makes something a rule system is some form of conditionals evaluated against a set of facts. Rules systems encompass everything from Prolog and OPS5 to first order logic. These systems have different syntaxes, expressiveness, and characteristic semantic techniques. Some are very declarative and (furthermore) KRy, and some are more like imperative programming languages. For RIF, what is desired is a common framework in which rule and fact sets in one formalism can be converted to those of another, preferably without loss when possible.

There are three main styles for such a framework:

1. Common subset + extensions
   Most of the formalisms build (syntactically at least) on Horn clauses. Unfortunately, the semantics differ quite a bit, but many can be seen as either a form of minimal model (or other preference) semantics or regular first order semantics. So, to handle something like Prolog, you would take the core (Horn clauses with minimal model semantics, maybe) and extend it (both syntactically and semantically) to encompass the features of Prolog. For this framework to support interchange between Prolog and and OPS5 like system, or Prolog and an FOL reasoner, it would have to say either, 1) hey, I can only help you if your Prolog programs are pure Horn clauses and you read them this way or 2) I understand how to map your extensions to the core into the extensions that suffice to describe OPS5 or FOL.

   But this requires a uniform semantics for all the extensions, or some combination methodology (a la fibring).
   

2. Superset
   Following on the need for a uniform semantics, we could try to describe an all encompassing rules language that would be a proper superset of all the rules systems of interest. Interchange is facilitated by determining if the translation into the background formalism of the rule set can be processed by the target system. FOL is sometimes proposed as the common superlanguage, but it has a lot of problems with, e.g., non-monotonic features. So, often, we’d be reduced to some axiomtization of or encoding in FOL, a la the LBase strategy. We could aim for a more expressive formalism, such as FOL plus a bit (e.g., MKNF or annotated logic).
3. Metalanguage
   Once you are in the land of encodings or axiomtization, that is, once you don’t have a one to one mapping (or nigh 1-1) of constructs of the Rules systems and the interchange language, you’ve moved closer to a metalinguistic approach. In essence, you could take the 1 strategy except that all the extensions have to be described in the metalanguage, which would have to be rich enough to usefully describe all that we want to exchange, and could show when two rulebases using variant extensions were actually equivalent.

1 and 3 are pretty much the same except you enforce a formalism for describing the extensions and use theorem proving to help out (rather than just using natural language and possibly no disciplined extension description, i.e., leaving it to each extension specification to specify, well, everything).

Alas, 2 or 3 (esp. 2) probably require a lot of research before we can know what’s practical. So, we’re back to writing extension descriptions and proving the requisite properties by hand.

One strange turn the Semantic Web Rules saga took is the elevation of Interchange as the point and purpose of the working group. If you go back to the prior round of discussion (which, in fact, resulted in a charter around the same time that the Data Access Working Group was being born) you won’t see “interchange”:

The group is chartered to develop a practical and useful rules language for the Semantic Web, along with a corresponding language for expressing justifications….

...This work is based the Resource Description Framework (RDF) in order to enable full interoperation with other elements of the Semantic Web, including the OWL Web Ontology Language and possible future languages for logic and knowledge representaion….

...The architectural vision for the semantic web has included an interoperable standard langauge for rules for a long time. The readiness and urgency in this field are growing, and have crossed the threshhold for Working Group creation….

Well, turned out not to be all that urgent, since it could wait for two years for any sort of group and the one we got was explicitly not chartered to do a Semantic Web Rules language!

(An aside: When interviewing for a lectureship at the University of Manchester, I was asked what one of my greatest frustrations was with my work. After a false start I said “Hype”. I’m really sick of hype. I remember being stung by hype back with the release of HyperCard. It’s wretched. I find it especially frustrating when I am seduced by the hypemonster myself. I think that negative hype is pretty obnoxious too.)

So, if the RIFraff are not doing a Semantic Web Rules language, what are we doing? What the heck is an Interchange Format and how does it differ from a rules language or a Web Rules language or a Semantic Web Rules language?

At the F2F we all agreed, properly, that whatever we produce will certainly be a language with a clear and well defined semantics. It’s the sine qua non of interoperability, hence of interchange. And we are chartered to produce a language compatible with RDF and OWL (in fact, one delieverable is a Rec track document describing OWL compatibility). So, what’s left?

To my mind, the main difference is in the desiderata and the design choices they impose. To hook in the rules/business rules vendors, the W3C has been trying to sell this as a language for exchanging rulesets between their systems. Whatever we produce has to be able to more or less do that. So, instead of starting from the thought, “What would we like a Semantic Web Rules language to look like?” we start with “What features must be supported by RIF in order to take rules from ILog’s thing to Jess?” Note that “webbification” is not, on this model, an inherent requirement. We’ll have it, but it’s not essential. Computational niceness is not an inherent requirement. Neither is stand alone useability. There’s a sense in which we are not supposed to pick and choose features…just to allow for two systems that support some set of features to be able to load common rulesets. (“Just” indeed!)

Given that we are explicitly working with a two-phase, core + extensions design, things are going to get complicated. Furthermore, there is the possibility of multiple semantics/semantic paradigms (e.g., as discussed in the OWL/RDF breakout session, we might specify that in this document implication is interpreted in minimal models or all models. More generally, it’s not clear that it will be easy or even possible (reasonably speaking) to provide a model theoretic/denotation semantics for lots of features/feature combination (leading to proposals to use formal operational semantics a la XQuery).

Illegal domestic spying, while itself not as bad as torture and extra-judicial indefinite detainment (among many other things), is still pretty amazingly awful. The exercise, even the petty exercise, of power without restraint should be completely unacceptable in our political culture. It is the antithesis of our democracy.

(Note, I will be including these little points at the bottom of all my posts because I can’t bear not to. I don’t have the energy to do the kind of political blogging I used to do at Monkeyfist, but I just feel sleazy posting nothing but on technical matters. I am thinking about this other stuff as I write these posts.)

(Title lifted, with irony, from B. Grosof’s proclamation at the end of ISWSC 2005 and the start of the RuleML Conference. Note that the “OWL workshop” (i.e., OWL: Experiences and Directions) was not “included” in the RuleML. We were co-located. I was an organizer so I should know. I do know. That bit of aggrandizement by Grosof is not too surprising, and fits with the proclamation.)

Or perhaps, My Year of Rules. Though there are like two years scheduled. Yes, the W3C has decided to start up a Rules Working Group. Yours truenfalsely is on it, and shall attend the first face to face, though I surely have better things to be doing. And worse things which need doing.

I am not a Rules Believer, and I get annoyed at them. Rules are not magic (neither are description logics) and Rules (note the magic caps) have their problems (including that the term is too generic, glossing over tons of useful distinctions).

I gave a presentation of my position paper for the Rules workshop, and it went over very well. I don’t know that we’ll see much like what I advocated, but there is Respect For OWL built in to the charter.

(I only want Respect For OWL because I’ve invested a lot in it. OWL isn’t magic either! I do think it has some advantages, e.g., object/class centrism, classification & realization, expressiveness; but some of those those can be weaknesses too.)

As I contemplated the dreary task of writing an introduction , impelled by the steady dripping of email bearing hopes, dreams, and use cases, I thought of one of my favorite quotes from the Tractatus (6.4312):

Not only is there no guarantee of the temporal immortality of the human soul, that is to say of its eternal survival after death; but, in any case, this assumption completely fails to accomplish the purpose for which it has always been intended. Or is some riddle solved by my surviving forever? Is not this eternal life itself as much of a riddle as our present life?

It fits all too well.