% notes from discussion of Rules at DAML PI Mtg 10/17/03
% by Benjamin Grosof, discussion moderator
% (note the remarks attributed to various people are paraphrases,
% not precise quotes -- apologies for any distortions introduced) 

o issues from discussion during OWL Rules presentation:

can extend via Lloyd-Topor to have disjunctions in antecedent

may be able to use a RuleML-based syntax, esp. with references to OWL

could have a more concise human-consumption string syntax;
may be able to use a N3-ish syntax for most/all of the expressiveness

empty heads:  how important to represent

perhaps relax to permit variables in consequents 
"across" class descriptions

needs study:  Horn + DL/OWL as KR -- its expressiveness, 
usefulness, and computational complexity 

implementation and tool approaches:  

tractable expressive subsets

sequence of development:  inferencing, then authoring and translation

implementability path is a criterion for choice of language

developing use cases and performance-scaling tests 

interesting combo:  rules + DLP restriction on class descriptions
- this is guaranteed tractable


o issues from discussion during RuleML Rules Lite presentation

needs study:  semantics of LP+DL, e.g., "DL-Augmented" Herbrand Model


please send feedback/comments to www-rdf-rules@w3.org 

Jim Mayfield:  please give us an immature something

Mark Greaves: 
we need tools, in addn to spec, to get a lot / deep feedback
from most users even in DAML

TimBL: we could use OWL Rules for just rules, for hybrid, can use DLP

Ian Horrocks:  
there's lots of recent literature on hybrid inferencing, of 
more general kinds, too

Katia Sycara:  START from use cases, for web-based reasoning:
what are most useful features and why
- some of the expository/theory-oriented examples don't grab,
(Benjamin Grosof: e.g., "don't cry uncle")

Ian Horrocks: actually the uncle example did address some use cases
 
Mike Dean: there's a use cases doc; didn't really ground the
discussion enough towards some key issues, e.g., binary relations
restriction of RDF

Mark Greaves:  were you happy with those use cases?
A by Mike:  not really; test cases were more useful than use cases
- use case is text that talks about what's desired;
test case is stuff that's more concrete and runnable

Mark Greaves:  my one major priority:  
- right kind of formalisms and syntax and mappings to support 
the needs of SW Services incl. composition

Mark Burstein:  it would be extremely helpful to separate the 
syntax issue from the rest of the discussion
- for Services what we need now is a consistent way to express things,
want a direction to get feedback about what we can and can't say;
i.e., be overly generous for time being, then can see which is
actually utilized in the Services 

Benjamin: e.g., slotted Situated Courteous LP in RuleML and OWL Rules
each have a fair amt of expressiveness, with slotted syntax in dev

Drew McDermott: I agree with Mark Burstein.  The current syntactic
proposal looks extensible to pretty much any of the expressiveness
(extension) directions being contemplated, i.e., that Benj presented.

TimBL:  yes, be permissive and name the expressive/syntactic subsets.
Rule applications tend be quite varied in what you need to do, 
i.e., what expressive features are used.  
We need a lot more experience.

Benjamin: One can view this as dialectic pendulum swings, very
possibly it's time for the swing back towards more expressiveness

Pat Hayes:  the important thing now is to have an inclusive syntax,
OK for someone (e.g., tool, application) to be incomplete wrt to that

Mark Greaves:  yes, I worry about coming up with the one true logic

Benjamin: e.g., consensus Horn-ish, richer in dirn of SCL/FOL, richer
in dirn of LP/RuleML, e.g., towards SCL and SCLP RuleML respectively,
with consistent overall RDF-y / OWL-integrating syntax and design
style

Bob Macgregor:  wanted also:  
- tutorial materials
- translator tools described in terms of shared conceptualization of 
  features

Mark G and TimBL:  let's have a set of named subsets of expressiveness
Benjamin:  that's the approach RuleML has taken all along

overall consensus -- summary:  
   (presented in afternoon joint DAML-OntoWeb rules session)

o Be ecumenical wrt extending expressiveness
  - Situated Courteous LP   &   FOL/DL 
  - experiment with needs
  - Horn case as strong-consensus, common
o Unified syntax ; that integrates nicely with RDF, OWL
o Use cases use cases use cases     test test test
  - Wide variety  ;  including Semantic Web Services
o Tools tools tools
  - CCI/RuleML engines, FOL engines  (Inferencing)
  - Translation;  Authoring