From: Benjamin Grosof ([email protected])
Date: 08/05/03
% notes from JC telecon 8/5/03
% by Benjamin Grosof
preamble: Mike Dean is also posting notes from today, as well;
his notes cover (unlike this by Benjamin) the first half hour
of the telecon which discussed comments on n-ary by Harold
agenda for today's telecon:
o comments on n-ary by Harold
o refine and consolidate overall agenda consensus formulation from last week
on what should be in an initial Rules Lite version
- including emails and feedback since last telecon
participants:
(see Mike Dean's notes)
%%%%%%%%%%%%%%%%%%%%%%%
o first half hour of the telecon, discussing comments on n-ary by Harold:
see Mike Dean's notes (Benj missed being able to record this)
%%%
issue to bookmark for later discussion:
naming of rules (and of OWL axioms)
%%%%%%%%%%%%%%%%%%%%%%%%%
o rest of telecon: discussing how to refine and consolidate the
overall agenda consensus formulation from last week
on what should be in an initial Rules Lite version
- includes requirements, expressiveness, work items, action plans, schedule
issue: restrict to Datalog? [Benj]
(discussed)
- Datalog here means no logical functions (beyond individuals)
-
Harold: what about permitting the RDF list constructor?
- consensus: yes, we probably should, for finite lists
- it's reducible to simple RDF without list
- OWL uses rdf:list but not rdf:seq
- discussion
- Pat: KIF/CommonLogic experience was
over-expressiveness/ambitiousness about lists, since KIF formulas
could be quantified over within the language, thus losing finiteness
- Benj: issue lurking of blank nodes being typically used in
axioms that employ RDF lists, and of this
enabling head existentials which are disallowed
- Pat: but this is probably reducible to simple skolemization, thus not
problematic; can further impose restrictions on what kinds of things can be
specified about such blank nodes, e.g., only at top level, have experience
with Peter about this kind of "infection prevention"
- more about this to be discussed on email
issue: restrict to disallow 0-ary relations? [Benj]
(discussed)
- consensus: probably, since simplifies the mapping to and from RDF
- Pat: can encode a 0-ary relation named p with
a single unary relation whose argument is named p
Benj: wrt equality predicate: we may want to restrict it in similar fashion
to OWL-DL, i.e., not as part of user-defined signature of predicates,
and not permitted in rule heads and possibly not permitted even in rule bodies
Ian: yes
(discussion)
Benj: want growth path beyond Rules Lite
issue wrt requirements: [Benj]
we'd like to facilitate use of Rules Lite and its future extensions
by systems that are not primarily oriented to RDF (or OWL), e.g.,
by existing relational database, Prolog, or Jess rule systems and applications
that use such.
Pat: issue of layering semantically on top of RDF
- Peter: let's avoid the kinds of problems/pain OWL had with this
- Benj: can view RDF triples
- (discussion about possible issues, e.g., treatment of variables in RDF)
Harold: possible issue of essential expressive limitation arising from
combination of finite lists with binary relations
- may not be able to show the general case of the usual reduction of
n-ary to binary
- more about this to be discussed on email
Pat and Peter: philosophy of "prevent infection" expressively
Pat: suggest addressing interaction with RDF and OWL up front
Harold: issue: want to deal with distributedness of facts and list members
****MAJOR DECISION: consensus is to ratify the formulation of Rules Lite from
last week
- binary Horn FOL with possibly further restrictions
- plus warning label -- which is important to state clearly and precisely
- timing: aim for first version, with running test cases, by time of
DAML PI Meeting and SWSI F2F in mid-October
Benj: let's aim for strawman version by halfway, i.e., say 2nd week of Sept.
**pieces to tackle in our overall Rules Lite effort:
requirements and objectives statement
expressiveness -- esp. what further expressive restrictions or additions
warning label
updated roadmap
semantics -- is simple since is just FOL essentially
abstract syntax
concrete syntax for exposition of examples, and for test cases
- ideally easily allowing use of current tools
examples, including test cases
running of test cases using various existing tools
summary of expressiveness issues, earlier discussed, that will need to
be resolved: [Benj, after the telecon, extracting from telecon discussion]
- Datalog restriction (probable)
- RDF lists permitted, but probably restricted somewhat:
. blank nodes -- avoid in head; perhaps also avoid in body
. avoid KIF's over-ambitiousness (consensus)
- avoid 0-ary predicates (probable)
- limiting equality as predicate
. in manner similar to OWL-DL (probable)
. esp. in head
- permit conjunction within head in rule syntax, using expressive reducibility
. naming of rules might be a complication
- naming of rules (perhaps defer beyond Rules Lite)
- how tight is layering semantically on top of RDF or OWL
- be able to handle distributed facts
some initial volunteering:
examples and test cases: Mike, Said, and maybe Deborah
warning label: Benj and maybe Pat
concrete syntax: Harold and Said
abstract syntax: Benj;
would also like later help from someone who worked on OWL abstract syntax
action plan:
for next week, all to please think what parts of initial Rules Lite
document etc. efforts each person will/might volunteer to help with
________________________________________________________________________________________________
Prof. Benjamin Grosof
Web Technologies for E-Commerce, Business Policies, E-Contracting, Rules,
XML, Agents, Semantic Web Services
MIT Sloan School of Management, Information Technology group
http://ebusiness.mit.edu/bgrosof or http://www.mit.edu/~bgrosof
This archive was generated by hypermail 2.1.4 : 08/05/03 EST