- From: Dave Reynolds <der@hplb.hpl.hp.com>
- Date: Fri, 26 Aug 2005 17:19:43 +0100
- To: Sandro Hawke <sandro@w3.org>
- CC: public-rule-workshop-discuss@w3.org
Hi Sandro, > You also observe that this seems like too much work. I'm wondering > what the best way is to estimate that. An actual list of things the > Working Group needs to do, and a guestimate of how much time each bit > will take? A useful thing to do, but perhaps better when there's more consensus on the requirements to be met and the broad approach. I guess my concern is not so much the sheer volume of things-to-do (though there is some of that) as the difficulty of squaring-the-circle on divergent requirements. From my (weak) understanding of business rule languages they seem to include clearly non-monotonic notions like default rules and can depend on the engine's operational semantics. A requirement to losslessly interchange such rules seems hard to square with beyond-OWL-KR. > What I think makes more sense is to have a charter where > at least of few members of the WG think the whole thing is a few weeks > work, and then the whole process can be slowed down as much as > necessary to polish and check the work, but as time runs out you stop > polishing and just call it done. > > I don't think the current draft has that quality, but I'm starting to > picture a two-phase version that does, where at least the stuff in > phase one is something you or I would feel we could hammer out in a > couple of weeks if we didn't need to reach consensus. Then saying the > consensus version takes on the order of a year seems more reasonable. Seems like a pretty realistic estimate of the cost of consensus! > Do you see a subset of this work that could done in such a timeframe > and would still be useful to some user communities? Not clear to me yet, sorry. If you slice the work down by picking one target community it seems more tractable. For example, the (vague) "RDF transformation and processing" area is small enough that something targeted just at that might be tractable but I don't think that would be useful for either business rule exchange or high level KR. > (I imagine the > answer is: what you have implemented in Jena. :-) I've tried to be very careful to separate what might be best for the WG from anything we happen to have lying around! > Have you done the > mental diff between that and the draft?) Clearly huge. We are a long way from FOL-with-equality, that's just nowhere near our design centre. Amongst other things we've dodged the hardest issues by only supporting negation over existential data. I guess conversion of Jena rules out to the FOL proposal wouldn't be so bad except for those cases where rules depend on our operational semantics, but you wouldn't expect to be able to interchange those anyway. There's quantification over predicates but hi-log syntax or representing rdf data by rdf(s,p,o) rather than p(s,o) would handle that. Dave
Received on Friday, 26 August 2005 16:20:32 UTC