Re: links & messages



Date view Thread view Subject view Author view

Joaquin Miller (joaquin@acm.org)
Thu, 20 Jan 2000 09:58:09 -0600


At 05:24 AM 1/20/00 , Andy Evans wrote: >At the most basic level, we model systems using objects and >links. I feel that is the best approach. I also feel it is crystal clear that is not the approach of UML. In UML, at its most basic level there are classes and associations. Objects are instances of classes. Links are said to be instances of associations. Links have a data content. The data content of a link is a tuple.* [I do not believe this approach is necessary. In fact, UML could be redefined, at the most basic level, to be about objects and links (and actions), and the principal effect on users of the current UML would be to make things cleaner and clearer.] As an example of the problems i see, consider these concepts, which are introduced without definition: actual association and real association. It makes one think. Maybe associations that are not actual are not relations? In that case, some associations could specify the existence of references. Well, after all, a set of tuples could be the set such that an object mentioned in the first place have access to a global value that identifies the object mentioned in the second place ("rather than an actual association"). Then our interpretation of the "association" could be that the "source" has a reference to the "target." Since we have a relation, is that an actual association? Or not? Meanwhile, is <<reference>> perhaps missing from the stereotypes Jos mentions (<<association>>, <<parameter>>, <<local>>, <<global>> and <<self>>)? Is isNavigable a proxy for <<reference>>. Is it fair, if not kind, to say the content of UML may still be a bit smeared across several categories. (If piling on were legal, i could ask: what is the (forgive me) relationship of reference to the concepts navigability and traversal?) (* The other content, the non-data content, of a link must then include the constraints such as the stereotypes Jos mentions.)


Date view Thread view Subject view Author view