W3C Workshop on
Rule Languages
for Interoperability
-
Panel: What's Next?

Jon Robert Pellant (Pega)

Here are some issues which I find could be beneficial outside of the obvious- customers demand standards to protect their rule assets from vendor lock-in.

1. Re-application of rulesets
   1. A rule standard must separate the binding of facts from the rule itself
   2. Dictionary publishing from author's perspective (OUT)
   3. Rule templates
2. Delegation
   1. Specialization
   2. Exceptions
3. Authoring leveraging semantic web
   1. Authors can define rules against existing dictionaries (IN)
   2. External rule authoring (KPI, BRS, BizRules, etc.)

Jean-Francois Abramatic (ILOG)

Q. What should we try to acheive?
A. Develop a rule language that allows the use of rules in multi-app, multi-platform situations

Q. What is the scope of the language
A. We should start with "simple" rules (see IBM's paper) in order to guarantee applicability, performance, and deployment

Q. What are the external constraints to the development?
A. The language should be compatible with existing standards (W3C, OMG) in some form.

Tim Berners-Lee (W3C)

1. Simple enough to be widely deployed; Make the minimal language but have a maximal language in mind as a unifying concept
2. Use URIs Seriously
3. Context-Free
4. Supports Partial Understanding

Donald Chapin (Business Semantics Ltd)

1. Make 'Community' the core notion for 'context' -- people using the Semantic Web need to identify what community they are interested in for this use of it.
2. Define a framework for rules, and the vocabularies they are defined in terms of, that extends from a fully business-optimized usage through the necessary transformations to executable rules.
3. Provide the ability to state that different expressions of terms, facts or rules all mean the same thing.

Kurt Godden (GM)

1. A tool-independent OWL-compatible syntax. To borrow a term from translation theorym I would call this an interlingua for representing rules.
2. First order expressivity, plus a wish list:
   1. lambda-binding to associate single domain entities to variables
   2. quantification over predicates so we could write rules/constraints over the alloowable types of properties
3. How about different flavors of rule languages a la OWL so that domain experts might be able to express simple rules themselves with Rule Light and work work a knowledge engineer for Rule Full.

Harold Boley (RuleML)

The three top-level items for a W3 Rules Language that were discussed as much as possible, in the given time, with WSMO, SWSL, and RuleML:

1. Kernel should be Datalog Horn + NAF;
   URI-identified syntactic and semantic extension layers (a Rules Stack refining the new Layer Cake) should include
   • Slotted/Frame syntax
   • Webizing as in RDF and N3
   • Reification
   • N3/TRIPLE/F-logic-like contexts/modules
   • Query sublanguage interoperating with SPARQL
   • Prioritized conflict handling
   • Production Rules (CA Rules)
   • Logical functions (full Horn logic)
   • Syntactic sugar for restricted higher-order as in Hilog and SCL
   • Human-readable (non-XML) presentation syntax
   • ...
2. Integration of Rules layers and Ontologies (DLP, AL-log, OWL-DL, ...)
   1. Lot's of previous work in RDF and WebOnt WGs to build on
3. Use cases drawn from Semantic Web Services
   1. Leveraging work on SWSL and WSML
   2. Policy Annotation
   3. Web Service Discovery
   4. Event and Process Modeling --> Extension layer for ECA Rules

Paul Vincent (Fair Isaac - Blaze Advisor)

1. Pragmatism: compatibility with other worlds (eg Object Oriented)
2. Role: rules for defining structure vs behavio(u)r (vs enforcement...)
3. Use: transmission (with data) across systems for the purposes of sharing & collaboration

Ed Barkmeyer (NIST)

NIST highest priorities

1. exactly one common abstract syntax with one corresponding XML exchange syntax
2. one or more well-defined interpretations = one or more semantic models for interpreting (explicit subsets of) the common abstract syntax, with rulesets marked as to their interpretation
3. must support basic datalog, n-ary predicates and constrained NAF interpretations
4. must support reference to (walled inclusion of) OWL ontologies and other rulesets
5. at least one standard surface syntax with mapping to the abstract syntax, explicit permission for other surface syntaxes provided that they are unambiguously mapped to the abstract syntax and a/the standard semantic theory
6. prefer a frame style of surface syntax with n-ary predicates

---

Sandro Hawke, W3C
$Id: panel.html,v 1.14 2005/04/28 21:08:20 sandro Exp $