SBVR — Section 1 - modeling based on MOF does not have all of the capabilities

The SBVR specification defines a metamodel that is not a MOF metamodel and then defines a transformation from the SBVR metamodel to a MOF metamodel for rules defined in the SBVR metamodel. This means that modeling based on MOF does not have all of the capabilities (e.g., reflective definition of vocabulary) of the SBVR model, and that MOF technologies (e.g. QVT) cannot be applied to the SBVR metamodel. Furthermore, this means that aspects of the SBVR metamodel cannot be integrated into other metamodel specifications so that, for example, business vocabularies (multiple) and business rules mught be specified for an organization structure model or a business process model.

Noun concepts are represented, in MOF terms, as classes, unary fact types as Boolean attributes, binary fact types as associations, and ternary fact types as classes. To be completely straightforward, this direct correspondence from SBVR meanings to MOF model elements requires two semantic modeling capabilities that are not in MOF 2, but which could soon be supported through a new initiative, Semantic MOF (SMOF). The two capabilities are support for multiclassification and for the open world assumption. Workarounds make the direct correspondence workable in the absence of SMOF as follows:
1. Multiclassification: When a thing is categorized in multiple ways such that it instantiates two or more MOF classes, neither class specializing the other, the thing is represented in MOF by multiple MOF objects which are linked together using an association to indicate that the multiple MOF objects represent a single thing (this approach to multiclassification is allegedly also being used in for ADM).
2. Open World Assumption: The open world assumption is that representation of facts in a model does not imply that those are the only facts of a particular type nor that they are the only facts of a particular type about a subject thing - there are no implications to be taken from what is unstated. MOF supports this open world assumption about instantiation of classifiers (classes and associations), but MOF's attributes support only representation of an entire extension of an attribute with respect to a given subject. In order to enable a clear distinction in a MOF context between individual facts and complete extensions, association links are used to represent individual facts of a fact type while attributes are used when identifying a complete extension of a fact type with respect to a particular subject. This means that a fact can in one model be represented using a link, and in another as a value of an attribute of an object.
The two workarounds, however clumsy, are used because they enable a direct use of MOF. It is anticipated that an acceptable SMOF will directly support multiclassification and the open world assumption, both of which are supported by RDF and OWL, and that SMOF will remove the need for the workarounds in the future.
Changes to the SBVR specification are summarized as follows:
1. Clause 5 explains that the MOF interpretation of the figures is explained in clause 13 (SBVR Metamodel).
2. The caption under each figure in the normative clauses 7 through 12 which says the figures are nonnormative is changed.
3. Clause 13 is entirely replaced by an explanation of the MOF interpretation of the vocabulary entries and figures, and of the direct semantic correlation of the SBVR vocabularies with the MOF-based SBVR Metamodel.
4. Annexes K and L are removed (they are no longer relevant).
5. Some fact type forms are added for existing fact types as needed where MOF attributes are required in order to indicate a complete extension of a binary fact type for a particular subject, especially where required by reference schemes.
6. Three signifiers that use characters (">", "<", "&") that do not work in an xmi:id are removed or replaced.
7. The use of xmi:id values in SBVR Metamodel documents is explained in clause 15 such that it supports URIs that refer to SBVR designations and fact type forms.
8. The use of square brackets around placeholders in UML figures is replaced by underlining so that the brackets are not thought to be part of an association's name.
The specific changes below assume that clause 2 will be corrected separately by the resolution to Issue 9957.