UML22 — UML2 should specify default property ownership for association ends

UML2 should define the defaults for property ownership and the explicit meaning of navigation notation. It should also provide notation for overriding these defaults in order to specify explicity the classifier that owns a property without relying on navigability. This recognizes a common notation practice and will result in predictable metamodel interchange. I general, the UML2 spec should attempt to avoid situations like the pair EF and IJ in figure 21, and instead use sensible defaults. If the default isn't what the model wants, then there should be notation to explicitly say what is needed. This will limit semantic variation points or unspecified notation meaning that may result in interchange and interoperability issues.

The diagram conventions used in Superstructure section 6.5.2 tie navigability and property ownership together in a manner that is consistent with the notaiton used in Basic and EMOF. However section 7.3.1 Notation for Association says:

Various options may be chosen for showing navigation arrows on a diagram. In practice, it is often convenient to suppress
some of the arrows and crosses and just show exceptional situations:
• Show all arrows and x’s. Navigation and its absence are made completely explicit.
• Suppress all arrows and x’s. No inference can be drawn about navigation. This is similar to any situation in which
information is suppressed from a view.
• Suppress arrows for associations with navigability in both directions, and show arrows only for associations with oneway
navigability. In this case, the two-way navigability cannot be distinguished from situations where there is no navigation
at all; however, the latter case occurs rarely in practice.

This is fine, but given a UML2 diagram what are we to assume if all navigations are not explicit as in the first bullet? Wouldn't such such a model be ambiguous? Should UML2 specify which one of these conventions are implied by the notation? The last bullet represents common practice as well as the conventsions used in the UML2 specification. Perhaps the UML2 spec should to be specific about what the notation means and not leave this up to the reader.

Later in the spec (page 42) under Issue 6243, Figure 22 shows a class containing a property with non-primitive type and indicates this is an ownedAttribute of the class, and can be shown as an association too as described in Basic and EMOF. What it doesn't say is what the notation

by itself means. We know ClassA can navigate to b, but we don't know anything about who owns the properties and therefore where the ends go in an instance of the metamodel. Are they both ownedEnds of the Association? Is b an ownedAttribute of ClassA and a is an ownedEnd? Since there is currently no notation for specifying which classifier owns the properties, the notation should specify the default owners. Otherwise different tools may produce different XMI as it is not clear when a property on an association end is an ownedEnd of the association or an ownedAttribute of one of the associated classes.

The conventions in 6.5.2 should be the definitive notation for navigation arrows (with x on the ends options to make non-navigable explicit), and also specifies the default for property ownership. That is, the bullet lists in 7.3.1 should be replaced with those in 6.5.2 for association navigability and property ownership.

Then a notation should be specified for explicitly stating property ownership when the default is not appropriate.

No Data Available