Joint Committee Minutes 30 October 2001

Participants

• Dan Connolly (joined late)
• Mike Dean
• Stefan Decker
• Richard Fikes (joined late)
• Frank van Harmelen
• Sandro Hawke
• Jeff Heflin
• Jim Hendler (joined late)
• Ian Horrocks
• Deb McGuinness
• Peter Patel-Schneider

Announcements

Last Week's Minutes

DAML+OIL Errors

RESOLVED: Jeff's proposed updates should be applied. ACTION (Peter): update files in submission. ACTION (Mike): update files on www.daml.org to reflect Peter's changes.

Submission

Dan Connolly needs a copyright release from every author (anyone who holds copyright). Peter hasn't received any yet.

WebOnt

The WebOnt home page (with links to charter, etc.) is publicly available here. Lists of members, etc., should be updated this week; invited experts may take a bit longer.

The first telecon is expected to be held in 1-3 weeks; the first F2F meeting will probably be in January

Jim thanked everyone (particularly Ora) for their help in making WebOnt happen!

Query

1. Existentially quantified objects include both RDF anonymous objects (aka bNodes, etc.) and inferred objects.
2. How do we indicate the number of answers requested? In replies, how do we indicate, e.g., "provably all".
3. To what degree should we check for equality among answers?

In Deb's thesis, an individual could be generated due to cardinality constraints, etc. Part of the motivation was that Classic doesn't fire rules until an individual is created (this may be considered a limitation of Classic).

Ian felt that explanations probably shouldn't be built into the language. He noted a classic example that typically surprises non-logicians: querying for people all of who's children are doctors will include people with no children.

Ian noted that mechanisms for returning inferred objects must somehow deal with the potential for an infinite number of answers. We agreed to discuss this separately.

We discussed the lifetimes of returned names. Classic stores names (identified as skolems) persistently in the Knowledge Base. Peter noted that skolemization is not always benign.

If we include information about skolemization in query replies, are we making assumptions about the sophistication of clients? We want to be able to support dumb clients (Visual Basic, perl scripts, etc.) accessing smart servers.

We might want DQL to support a contruct like SQL's select count as well as a special case for determining existence (any, count >= 1).

Richard and his graduate student Yulin Li are making progress on DQL. They expect to have a draft of a more specific proposal shortly.

Frank suggested that it would be good for someone to develop a summary of the query language issues raised in our email and telecon discussions. Richard volunteered, once he returns to Stanford from a family emergency in Texas.

Sandro asked which email list should be used for query discussions (joint-committee or www-rdf-rules, which includes query in its charter). The consensus was the joint-committee should be used for on-going discussions that required our shared context; self-contained proposals for wider comment can go to www-rdf-rules (with a link sent to joint-committee).

Next Week

Links