Joint Committee Minutes 17 April 2001

This page summarizes the telecon for the Joint US/EU Committee on Agent Markup Languages held from 1300 to 1400 PDT on 17 April 2001. These minutes were prepared by the chairman, and were approved during the 24 April telecon.



We've switched teleconference operators to Sprint, because Verizon will no longer give us both 800 and non-800 access numbers.

Last Week's Minutes

The minutes from April 10 were approved by those present.


Stefan distributed a message prior to the meeting to seed the discussion. Most of the discussion centered on whether/how we should have subsets/layers, rather than specifics of his proposal.

What would the subset address, e.g. an interface easily usable by simple Java and Visual Basic programs? Peter discussed the "canonical least model property", where one can essentially precompute all conclusions and require no run-time search, by eliminating disjunctions, etc. This is roughly equivalent to the "unique minimal models" discussed, for example, in Hector Levesque's thesis. This approach is used in Classic but often runs slower (without "amortization") than DLP on most real examples.

Subsetting can be confounded by various "backdoor tricks", e.g. minCardinality can be used to achieve disjunction.

Stefan, Frank, and Deb have received informal feedback (including from commercial companies) indicating that there is a significant niche between RDFS and DAML+OIL. Stefan's proposal was based primarily on this feedback. He isn't yet dealing with these issues in developing Triple.

Deb noted that Jim had earlier put a high priority on limiting the "conceptual load" of DAML. Is this still a major concern?

Frank noted that, although On-To-Knowledge has committed to using DAML+OIL, responsibility for the inference engine is a "hot potato" because of its perceived complexity.

Pat noted that suggesting "stepping stones" requires a lot less effort than definining named language subsets. Such subsets might be more worthwhile once we have a larger community (e.g. KIF is just now doing this).

There was some discussion over whether DAML Lite should be viewed as a subset of DAML, a superset of RDF, RDF(S) 2.0, or something else. Does the term "subset" stigmatize implementations?

Frank has been looking at feature usage in the DAML Ontology Library. The DAML program homework assignments, which comprise most of the current library, aren't necessarily representative of DAML usage, due to due dates, learning curves, etc. Also, advanced users are frustrated by the lack of rules in DAML.

A consensus began to emerge in favor of developing a DAML style guide rather than defining subsets at this time.

ACTION (Stefan): evolve message into a style guide, which Peter volunteered to review.

ACTION (Deb and Stefan): collaborate on identifying user needs.


Pat referred the DAML+OIL imports mechanism to the IEEE SUO KIF group. They believe that We'll continue the imports discussion when Dan Connolly is available to participate.

Next Week

We'll plan to discuss daml:imports.


Kelly Barber's raw notes
last week's minutes
Joint Committee home page

$Id: 2001-04-17.html,v 1.8 2001/05/04 18:53:01 mdean Exp $