Joint Committee Minutes 23 January 2001

This page summarizes the telecon for the Joint US/EU Committee on Agent Markup Languages held from 1300 to 1400 PST on 23 January 2001. These minutes were prepared by the chairman, but have not yet been reviewed by the participants.

Participants

Announcements

John Flynn sent a message to European participants about calling cards, etc. Please contact him, if you haven't already.

www.daml.org will down on on Thursday, January 25, from 5-6pm EST to move to a new server and IP address.

Last Week's Minutes

The minutes from January 16 were approved by those present.

Concrete Types

There was a lot of discussion about the revised proposal from Ian and Peter.

Santa and his associatedData in the revised example shows the use of an "any" or "top" type whose actual type (any concrete type) is declared when the value is instantiated. This is useful, but it's not clear that XML Schema has a corresponding mechanism. Also, the example shows the property as untyped -- perhaps we should have an explcit type indicator such as "top".

Jeff questioned the need to use different parallel properties (e.g. hasDataType/hasClass) for restrictions on concrete and abstract classes, increasing the size of the DAML namespace. Peter indicated that reusing the same properties probably wouldn't complicate the semantics, but they have different ranges.

Tim expressed a desire for properties that could take on either concrete or abstract values, e.g. a title that could be a string or an object containing multiple translations of that title into different languages. This conflicts with the disjoint-ness of concrete and abstract types in the current proposal. Mike suggested adding a higher level "top" that could allow either concrete or abstract types. ACTION (Tim): Provide examples motivating the ability to mix concrete and abstract types.

There was some discussion of including a set of boxed classes (analogous to java.lang.Integer) as part of a "standard library" accompanying the DAML language. Each would contain a value of the corresponding concrete type.

As part of a group effort to achieve closure on concrete types, Deb made a motion that we accept the current concrete types proposal at next week's meeting. ACTION (everyone): review the current proposal, post any counter-proposals to www-rdf-logic@w3.org, and be prepared for final discussion and a vote. Counter-proposals may include specific changes to DAML+OIL(+CONCRETE) and/or specific mechanisms or documentation to allow for future language extensions.

Language Change Impacts on Tool Developers

Deb mentioned that Jessica Jenkins had to add additional and rather complex code to her Ontolingua DAML parser as a result of the decision to drop restrictedBy in favor of the semantically equivalent subClassOf in DAML+OIL. Ontolingua has different mechanisms to represent subclasses and restrictions. ACTION (Deb and Jessica): post a more complete statement of the problem and issues to www-rdf-logic@w3.org.

Qualified Restrictions

Ian presented an example (subsequently posted to www-rdf-logic@w3.org) that clarified the need for hasClassQ, cardinalityQ, etc.:

If we want to say that a Person can have at most 10 children, we could write

<daml:Restriction>
  <daml:onProperty rdf:resource="#hasChild"/>
  <daml:hasClass rdf:resource="#Person"/>
  <daml:maxCardinality>10</daml:maxCardinality>
</daml:Restriction>
If we want to also say that at most 3 of those children are Doctors, we could add
<daml:Restriction>
  <daml:onProperty rdf:resource="#hasChild"/>
  <daml:hasClassQ rdf:resource="#Doctor"/>
  <daml:maxCardinalityQ>3</daml:maxCardinality>
</daml:Restriction>

If there were no distinction between hasClass and hasClassQ, this would *also* say that there are at most 3 doctors overall, which is not what is intended.

It would be desirable to put an example like this in the walkthrough.

RuleML Interaction

Harold Boley expressed interest in meeting with us to discuss DAML-Rules at the upcoming DAML PI Meeting. Mike referred him to the minutes of our 2000-12-07 meeting and expressed his belief that mid-February is probably premature, although we look forward to working with them in the future. Benjamin Grosof and Drew McDermott are involved with both RuleML and DAML.

Next Week

Planned topics for next week:

Links

IRC
last week's minutes

$Id: 2001-01-23.html,v 1.7 2001/01/25 07:58:02 mdean Exp $