SBVR — Section 2.2.1.1.1 Fac Type: "concept1 specializes concept2"

Fac Type: "concept1 specializes concept2"

This definition of this fact type includes the following example:

"The role ‘sum’ specializes the noun concept ‘number’, the differentiator being that each sum is the result of adding up a set of numbers. It turns out that every number is a sum of that number added to zero, so the extensions of the concepts ‘sum’ and ‘number’ are always the same."

Issue on: "the extensions of the concepts ‘sum’ and ‘number’ are always the same"
The extension of a role is not a set of things but a set of involvement of things. In extension of sum example, it is not a set of numbers but a set of an involvement of numbers
3 + 2 = 4 implies that 4 is involved in sum with an involvement of 3 and an involvement of 2
2 + 2 = 4 implies that 4 is involved in sum with an involvement of 2 and an other involvement of 2
Therefore, the extension of the "sum" role is:

.

In version 6 of the SBVR specification, fact-type role was changed to "role", which is now a direct specialization of "concept". To express what concept can range over a role (play the role), the only available mechanism is concept specialization. Hence the example provided by Don:
wife
General Concept: woman

Concept Type: role

We should be able to specify the role of "wife" in the context of marriage with husband without specializing it from woman or even from person.
We should then be able to define the range of concepts to which the wife and husband roles can be applied: "woman/men", "female/male" or in a Disney movie, like Robin Hood, "animals".
I don't see the value of being forced to go to the hierarchy of concepts (specialization) has a fact-type between role and noun-concept.

Previously, a fact type role was an object type. But it is the nature of object types that two object types that incorporate the same characteristics are the same object type. Roles of fact types do not have this property. A role of a fact type is intrinsically tied to a point of involvement in a fact type, and in the case of invertible fact types like 'thing is thing' and 'person is married to person', and in some other cases, two fact type roles can incorporate the same characteristics, or a fact type role can incorporate the same characteristic as an object type. Also, an individual concept is not an object type because it includes the sense of there being one instance.
The resolution of this issue makes the following changes:
What was previously one concept termed 'noun concept' and 'object type' is split into two concepts. 'Noun concept' keeps its original definition, but 'object type' is changed to be a specialization of 'noun concept' that is intrinsically based only on incorporated characteristics and that excludes fact type roles. The necessity that was given for 'noun concept' goes with 'object type'. 'Fact type role' is then intrinsically tied to a point of involvement in a fact type such that its incorporated characteristics are those understood from the fact type.
'Object type' is then understood to be what ISO intended by its 'general concept' and is one specialization of 'noun concept', with 'individual concept' and 'fact type role' being two other specializations.
With these changes it is possible that a fact type role ranges over an object type that has exactly the same incorporated characteristics as the fact type role. For this reason, 'specializes' is not adequate to describe that relationship, so a "ranges over" fact type is added.
The extension of a role is the totality of objects to which the role corresponds (as defined by ISO 1087-1). E.g. the extension of a role 'husband' would be all of the husbands. However, Antoine is correct that is it not necessary that a role be defined to specialize any particular concept or to range over a particular single concept. A note is therefore added to explain that roles need not be so constrained. Also, the range of a role, if given at all, can be given as coming from many concepts (an anonymous union). Also, advices of possibility can be used to indicate concepts whose instances can fill roles.
The unresolved "role playing of thing in occurrence" and "involvement" have been submitted as two spin-off issues.