PRD High Priority Items

From RIF
Revision as of 21:56, 12 January 2009 by PaulVincent (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Back to PRD Discussions Page

This page is a dedicated to list items with the highest priority after WD2.

From Christian

From Changhai Ke, ILOG

RDF / OWL interchange

Make sure that the inheritance of Core will give PRD the true capability to interchange with RDF / OWL. Wish to see real examples. The interchange with RDF / OWL is a major objective for PRD.

Basic level of rule interchange (level 1)

Make sure that PRD can represent easily simple rulesets that:

  • Use a simple model, flat, typed, with or without inheritance, but with no methods (the model is sometimes known as templates, or internal objects in some rule engines).
  • Uses some basic rule constructs, to be determined.

This basic level is autonomous to express simple rulesets, which are very useful for discussion purposes, for testing, or for benchmarks.

This level is the right one for model checking: the operations performed on the data are perfectly transparent in the rules, so the model is checkable.

If the simple object model can be assimilated to the frames, then we do not need to add new one. For this, we need concrete examples to be sure.

Industrial rule interchange (level 2)

This level targets to interchange real rule bases represented in the various rule engines. It should include:

  • A way to represent OO models, with inheritance and polymorphism, methods, etc.
  • A set of rule constructs, more extended than the basic level. This remains to be scoped.

This level is mainly designed to favor interchange between rule engines, which have similar concepts. It should not be disturbed by irrelavant concepts or considerations from other dialects in RIF.

Others

  • Support of XML schemas (as object models) for level 2.
  • Review of the condition formulas, and maybe enrich it with some new constructs (for example a not).
  • Extend level 1 condition part to level 2, add other constructs.
  • Determine the action part, for level 1 and level 2.
  • Pursue the operational semantics.
  • Build a first version of the XML schema for PRD.

From Paul Vincent, TIBCO

  • Objects vs Frames and effect on RIF.
  • Scope of PRD vs PRD Dialects.

Objects vs Frames

W3C RIF has targeted model-based frame-oriented fact representations (/tests). A major question for PRD (and possibly other dialects too) is how RIF should interact with other fact representations, and in particular XML.

One issue for PRD is that if PRD is to become a de facto rule exchange standard, is needs to be compatible with (and be comfortable to users of) the main data / document exchange standard, XML. This means not just mapping XML document data to facts for processing, and mapping frames to and from XML schema, but also an XML representation for PRD so rules can be embedded and exchanged in XML documents. Use cases for this are the main XML DSLs such as RosettaNet, ACORD, XBRL etc - which will need to use PRD rather than some other interchange mechanism if PRD is to be successful.

Note that "waiting for the domain document community to adopt semantic web representations such as RDF" is not considered reasonable!

Possible solutions are:

O vs F: RIF4DD

Name: RIF for Domain Documents (working name)

Description: Alongside RIF PRD provide a semantically equivalent "dialect" suitable for embedding in XML, with rules referencing XML content. Possibly the latter could use XPath.

For: Meets goals of RIF authors and potential end-users.

Against: Standards versions are standards dilution. Which would be supported?

O vs F: RIF-XML translator

Name: RIF XML translator (working name)

Description: Provide a common XML document-to-frame utility and semantics for use by all RIF dialects.

For: Simplifies RIF solution?

Against: Unlikely to encourage RIF adoption by domain community. At runtime RIF documents will instead be converted to native engine formats making which adds an extra translation layer (XML to RIF, RIF to BRE).

PRD vs PRD Dialects

With most PR engines having small variations in rule format and semantics, it is possible that RIF PRD would splinter into dialect-per-vendor. If so, it might be a "natural standard evolution" for these to be developed and then brought together in a later version of PRD...