Why do we want RIF?
(Semantic Web Perspective)

• A companion to OWL -- another logic
• Can express difference kinds of things
• A different way of thinking -- which many people want
• Good for data integration, ontology mapping
• ... maybe even data processing (it's a style of programming)

Why do we want RIF?
(Business Rules Perspective)

• Business Rules: thinking in the terms of the business
• reduce vendor lock-in
• coordinating businesses, eg in a supply chain
• integrated decision support

SW-BR split is a challenge

• should fade in time
• but we'll still be left with: industry vs research

Who is involved?

• Business rules vendors (ILOG (co-chair), Fair Isaac, Oracle, TIBCO, JBoss...)
• Commercial Research (IBM (co-chair), HP, ...)
• Non-Commercial Research (DERI, MITRE, REWERSE, ...)
• ... others (Betfair, OMG Liason, ...)

The Story So Far

• Workshop, April 2005, 80+ people
• Working Group F2F1, December 2005, 50+ people
• Use Cases and Requirements, WD1 March 2006
• Working Group F2F5, Feb 2007, 15+ people (comfortable size)
• RIF Core, WD1 March 2007

Two-Phase Charter

• The overall challenge is large.
  • Many different kinds of rule systems (Prologs, F-Logic, Production Rules, RDF Rule Languages, Database Rules, Event-Condition-Action, ...)
  • Long traditions, diverse communities
  • Little experience with W3C
  ... so Two Phases.
• Phase One: Define a common core, the set of features nearly all rule systems implement.
• ... but this RIF Core is not big enough to cover most needs
• Phase Two: Add extensions, based on demand, to cover clusters of systems
• ... but we're still in phase one.

A Brief Look at RIF Core

• Horn Rules
• Integration with RDF
• Extensibility

Horn Rules

• The relatively-easy part of the territory
• if condition then condition
• means: whenever the first condition is true, the second one is also true
• no negation: can't say "if X is false then ..."
• no actions: can't say "if X is true do Y"
• second condition is atomic (can't say: "if a then (b or c)")

Example Horn rule (in made-up syntax): two people with the same name and homepage are the same person:

If { ?x rdf:type foaf:Person. 
     ?y rdf:type foaf:Person.
     ?x foaf:name ?n.
     ?x foaf:homepage ?h.
     ?y foaf:name ?n.
     ?y foaf:homepage ?h. }
then { ?x = ?y }

RDF Integration

This work is underway. Not all of it is reflected in latest WD.

• RIF uses IRIs like RDF, to name individuals, predicates, etc
• RIF uses XML Schema Datatypes like RDF
• RIF is expected to support RDF triples as well as n-ary predicates, functions
• RIF is expected to support some XQuery/QPath Functions and Operators, like SPARQL

Extensibility

• RIF Core defines a "forward compatibility" mechanism
• ... so new features can be received and handled appropriately (ignored, force a warning, etc) based on metadata about that feature
• Likely Extension: Actions
  • if P is true then invoke-this-method
• Likely Extension: Negation
  • classical negation: if P is false then...
  • negation-as-failure: if P can't-be-proven-true then ...
• Possible Extension: Uncertainty Reasoning (new XG)