September 25, 2001

Attendees:
Harold
Frank
Jeff
Ian
Mike
Pat
Stefan
Richard
Ora

Announcements:


Agenda:

minutes from September 18 - approved

query

  RDQL (Frank, Mike, et al)

       Richard: This isn't appropriate for us.  It doesn't have the
       reasoning that we need.

  DQL and RQL (Richard) [2]

  Would expect that the query languages for RDF and DAML would be not
  entirely different.  Perhaps stacked like the languages themselves
  are.

  Richard: distinction between a standard database query language and
  what we are creating is that reasoning is involved.

  Jeff(?): our language will not be restirctive, but permissive.
  Another big difference.

  Information Exchanged During Query-Answering [3]
  
  Some questions from emails.  

       Frank: What if reasoning is useful, but does it belong in a
       query language?

       Mike:  can be useful.

       Richard: This is such a common type of reasoning it should be
       included.

       Mike:  what other systems have done this?

       Richard:  First Order logic theorem provers can do this. 

       Pat: agrees with Richard.  RKF: the kind of questions people
       ask a kb have this hypothetical feel.  But perhaps there is a
       difference between a query language and a question asking
       language.

       Mike: With hypothetical reasoning (what ifs), do you usually
       have two separate stores, one for the new facts, or do you add
       the new ones to the original store and then take them out
       later?

       Ian: isn't this basically subsumption style reasoning like you
       would have in a description logic type system?


       Richard: are we going to allow disjunctions and negations in
       the query language?  I would welcome input on that.

       disjunction is the more harmless of the two ...

       Richard: asks Ian to look into issues of decidability if we add
       negation and disjunction.

       Ian: I will look into it.  May not make it undecidable, but
       much harder for sure.

       Mike: the issue of returning 1 (or 10) instead of 1,000
       instances of things. (?)

       Also the issue of duplicate names for the same objects.

       Also the issue of if you want the first answer, or all of them,
       or all that you can get in 2 seconds, etc.

RDF Core WG (Pat, Mike, Ora, et al)

  has officially dropped prohibition against cycles in subClassOf and
  subPropertyOf

  Is "daml:collection" a string or a QName?  If the latter, to which
  namespace does it refer?

  conversation deferred til next week

Telecon Scheduling - what would 2 hours earlier do?	
	harder - I
	easier - I
	same - I

	I think that there is no consensus about this.  Let's solicit
	input on this and stick with the current time for now.


next week

  Richard - Query language discussion continuation

  RDF Core - daml:collection issue