[ontoiop-forum] Call for final comments on DOL, UML

[Bock, Conrad]

Alexander,

 >  > [Conrad] "Not sure I'm reading this correctly, but it seems to say
 >  > the properties involved in a composition association are
 >  > properties of m, whereas they are usually properties of the
 >  > classes at the ends of the association.":

 >  [Alexander] It is not intended to express ownership here.

Thanks.

 >  [Conrad]"Properties can be composed without being member ends of an
 >  association.  The UML 2.5 spec describes aggregation in the Property
 >  subclause (9.5.3), without constraining the properties to be member
 >  ends of associations.  The phrase quoted here describes a composite
 >  property with ("grouping its") values.  Feel free to ask Ed and
 >  others, this is well known to be the interpretation of the spec, and
 >  is implemented in tools.":

 > [Alexander] Yes, indeed, a property itself could be composite without
 > being a member end of an association.  I only now get the semantics
 > of such a declaration, say, c.p : \tau'[c'] with p composite (if I
 > understand correctly): it would be that if o is an instance of c, and
 > o.p are the values for property p, these values are parts of o (and,
 > hence, deleted if o is deleted - which we do not discuss); each such
 > value can only be contained (as a part) in a single object.

That's right.

 > Currently, we only discuss the case where a composite
 > property is a member end of an association, but with this
 > clarification we could also handle the "non-member end" case by
 > adding to the definition of "attributes" that p may also be decorated
 > by a filled lozenge and adding the ownership semantics.  In any case,
 > the text starting with "In UML, each Property may have
 > AggregationKind composite" should either clarify that we only discuss
 > the "member end" case or could be skipped altogether.

The first solution would be better if you have time, since it covers
more of UML, but if coverage of composition is left restricted to
associations, it would be enough to remove the text saying that UML
limits composite properties to be member ends of associations.

 > (The quote with "grouping its" is now on page 110 of the UML 2.5
 > specification, the constraint for member ends of binary association
 > on page 218.)

The text comment on the constraint should be "The only associations that
may be aggregate are binary", which is what the OCL says.  I can see how
the current text could be read as restricting aggregation to
associations.

Conrad