Coverage of Use Cases in RIF UCR 15 Feb 2006
(from the straw poll)

There is a general concern that many of the use cases need work to emphasize those elements of RIF use and capability that are perceived by the authors/supporters to be critical to the success of the specification. It may be that several of the coverage issues can be resolved by making such changes.

The notable cases of inadequate use cases were:

· Information Integration, which actually deals with the integration of rulesets from multiple sources. Two techniques are possible: importing the different rulesets into a single engine and processing them in a uniform manner, which uses the RIF for export from the distributed ruleset repositories, or accessing the rulesets by querying remote engines and processing the results, which seems to be out-of-scope. (Many)

· Message Transformation, or "Cross-Ontology exchange", which involves the transformation of information between two knowledge bases that use different ontologies via a third party agent whose transformation rules must be made visible to one or more of the participating agents. The main thrust here is the specification of transformation rules that operate on ontologies captured in RDF or OWL. But the use case doesn't indicate what special relationships that rules language would have to have with RDF and OWL, or what special features the RIF would need in order to support it. It should be noted that an accepted ruleset for mapping a subset of one commonly used industry ontology to another can be the basis for an expanded ruleset covering a larger part of those ontologies, or the basis for a conversion ruleset for a derivative ontology, e.g. a company-specific one. And this reuse of "mapping rulesets" creates the requirement for a standard exchange form. (Reynolds et al.)

· Rich Knowledge Representation, which doesn't refer to many of the KR features discussed in the corresponding Wiki section, such as frame-based models, higher-order features, etc. In general, the use cases for interchange of knowledge represented by RIF rulesets and knowledge represented in other well-known classes of repository are not very clear on the representation capabilities that would be required. And further, the relationships between "execution" of RIF rules and inferencing in associated KR systems need use cases that indicate the most significant requirements. (Kifer et al.)

The more serious concerns relate to concepts that are not addressed in any current use case. Some of these may be considered out-of-scope. The missing use cases are the following:

· Relationship between RIF and reference models (ontologies, data models, object models) for the terms used in the rules, which includes at least the identification of classes and properties and value sets and named instances. (Nichols, deSaintMarie)

· Relationship of RIF to the publication requirements of "webservices" (and in general to any collection of distributed functions). This is about what functions the service performs: rules that describe the relationships between pre-conditions and post-conditions, constraints that relate to the service parameters, etc. (as distinct from, or in addition to, process and choreography rules). (Nichols)

· Meta-information about rulesets, such as subset languages, use of standard patterns, derivation from other forms (such as OWL), and relationships to other rulesets. (Malhotra)

· "Vendor-neutral persistence" -- the use of RIF as an "archival" format for rulesets, that may be retrieved as needed for business processes and used by whatever engine is at hand. (This may just be cross-platform requirements.) (deSaintMarie)

· Actions in the consequent -- the invocation of functions that actually perform tasks other than modifying some knowledge base or providing a recommended action to a human user. In particular, functions may perform physical tasks, commit to "information actions" outside the scope of the engine (like contractual changes or financial transactions), or modify external knowledge bases, and may in so doing affect the perceived "state of the world" in which the rules engine is running. (deSaintMarie)