Attendees:	
Lynn
Deb
Mike
Ora
Jim
Ian
Kelly
Stefan

Announcements:
We got a nice response to our letter to RDF_core group


Agenda:
Minutes from last week .... no updates. 

Rules:
Mike chatted with Ben Grosof

Jim.  Differences in opinion of what rules are / mean / include?

Jim: More of us have played more with the 'framish' side ... need more
examples and documentation with the ontologies, etc

Q from Jim.  If there will be a web ontology / rules group does this
group want to continue to push DAML+OIL?

Deb: we need to present ourselves to the community better, will
generate more excitement on DAML+OIL ...

Jim: look at tools, etc, beyond ontologies.  Logic, rules, etc.

Jim: people say they can't do much interesting without rules.

Ian: pull in opposite directions?  some people - rdfs is basically
enough.  other people - need super expresive / bigger?

Jim: no, mainly people just need rules but the basic thing is good.

Mike: you can express those basic rules on their own.

Jim: need applications of DAML+OIL more usefully (not just to design
things to show off the ontologies / language).
    
Jim: how do you use this ... not just killer app sense.

Jim: instances from our documentation aren't easy ...

URL from Jim (http://www.cs.umd.edu/~hendler/jhendler.daml)

~~~

Jim: seems like we have done lots of TBox and not much ABox
.... (termination vs assertion)

Jim: if I'm a mom and pop shop and I add an ontology ... I can use all
these great tools ... then I will do this.  I won't be able to find
one ontology that has everything I need, but maybe I can use a module
or modules to make up what I need.  

We need to have stuff that says equalTo etc.  

Jim: If I create a person with things from two different instances
(people) and I want to say that this guy was created from person a and
this one from person b ... and by the way I mean these two to be the
same person ... what does "mean these two to be the same" mean?

Deb: Here's an example we think shows good form.  Here's a
recommendation as to what we think is important.

Jim: its not just examples but maybe new linguistic ideas ... same as,
kinda same as, really not same as... how do we do this?


Stefan Joins.  Rules conversation really starts ...

Two responses on Stefan's request for info on rules. Jim and Mike.

Rewriting / Transformation, Matching, Equivalence, Computation Rules, Class Rules, Trust Rules, Proofs, etc. are the ones that he
got.  We need more data to sample from before we can say much about
this. 

Jim: Can we get some background to make sure we all mean the same
thing.  What's the difference between a query and a rule.  Jim / Deb
.... select X from Y is not a rule. Stefan, absolutely is a rule.

Deb:  let's get a definition. 

Jim: do we mean rules as in creating new information from what is
already there.  Or do we mean rules as in rewrite rules?

Stefan: perhaps we just start with Horn clauses, negation, information
models, ect, and not specify (you can make rules to do inference
rules, queries, rewrite rules, graph transformation, etc.  Joint
infrastructure essentially.  What foundation are all of these based on
that we can extend.

Ian: we can do this (x is minor if x is person and person.age less
than 18) with DAML+OIL (if we depend on XML datatypes).

Stefan: Whatever rule language we come up with will inevitable have
overlap with DAML+OIL

Jim: is this a way to ground our language, but is it an ability to do
something new with the language that you can't do without rules?

Stefan: in order to do graph transformation ... we need the concept of
context.  (similar to a comment ... triples become quadrupiles).

Ora: don't most rule languages do that? 

Stefan: yes, because it is necessary

we ARE talking about syntactic closed world, not inferential closed
world.

Seems we're talking about Negation as Failure ... everyone is ok with
either versions of closed world, as long as the triple not being found
in the set (either as fact, or as inferred), does not give me license
to do anything, we are ok.  As soon as its lack of presence gives me
license to do something, they we have problems. Is X part of the
subgroup of people with no children?  In my set of triples, I have no
triples of (x, child, y) so it failed.  If I assume that X has no
children, we are in trouble.  If I just say I don't know that X has
any children, we are ok.

Jim:  would like the rules to NOT be tied a specific layout, etc. 

Lynn: with only affirmative rules (no negation as failure), but we
need defaults!

Deb: usually tricks that allow you to make defaults ... so I can get
by in a language with no defaults (only affirmative rules).

Ian: what are we really talking about?

Could default rule be another example for Stefan's list?  (completion
rules?)

Lynn: how about instead of doing THE thing, but rather A thing, and we
can do another thing later. 

Jim: A horn like representation attached to DAML+OIL will be very
useful, say most people I have talked to.

Jim: Does the expresion of a rule imply that the two sides exist. F(x)
-> G(y) ... does that imply that there are F(x)'s and G(y)'s that
exist?

Mike: Tagging vs Reification?

Ian: Pat's proposal, what was wrong?

Stefan: Context was missing.

Ian:  context?
      
Stefan: source is one example. 

Ian: will that have semantic implication, then why is it part of the
rule language.

Kelly: got lost in the technical discussion a little ... missed some
of it.  :)

.... break from rules ...

Things important for tomorrow.  

....Solicit comments and feedback 

....go into it with a feeling of what the role is for the committee
      
      Jim: feels its good to keep going, but if lots of us are in
      other groups, then we may be duplicating effort ... good for us
      to give a recommendation to the standards group.

      

.... for the future...

discussion of interaction between multiple ontologies