Re: Dependencies and associations



Date view Thread view Subject view Author view

Joaquin Miller (miller@joaquin.net)
Fri, 22 Jun 2001 11:28:57 -0700


Stuart Kent wrote: >The challenge that the UML 2 RFP has set is to find a language definition architecture which supports the development of a family of languages along these lines. I think this is a difficult challenge. Any views on this? I feel this is a very difficult challenge. I also feel that, if we can define a concept exactly in the UML 2.0 kernal language, that will help to define a related concept in a member of a family of languages based on that kernal. It ain't gonna be easy, but: If each relationship type in the kernal is specified using an invariant, then a specialization of that relationship type can be specified by adding to that invariant (for example, a specialization in a member of a family of languages based on that kernal, or in one of several viewpoint specifications of a system, or in a specific project or homework assignment). For those not familiar with the standard general relationship model (X.725), a relationship is defined by GRM as a collection of objects together with an invariant referring to the properties of the objects; a relationship type is defined as a named set of relationships sharing the same definition. [Yes, that is unabashedly West Coast style. Send any complaints to Geneva.] When that job is done, we can proceed, as advised by GRM, to a separate task, and specify how a particular relationship type will be implemented (perhaps by a specified relationship implementation type (or more than one such type); perhaps according to some generic relationship implementation type (pattern); perhaps following one of the GRM generic relationship representations, for example, "by means of participant-pointer attributes exhibited by the participants in the ... relationship."). PGP Fingerprint: CA23 6BCA ACAB 6006 E3C3 0E79 2122 94B4 E5FD 42C3


Date view Thread view Subject view Author view