Sparx Systems Forum

Discussion => Uml Process => Topic started by: Dennis L. Hughes on June 10, 2008, 05:44:08 am

Title: 4+1, packages, and models
Post by: Dennis L. Hughes on June 10, 2008, 05:44:08 am

I gather I'd like to use the 4+1 view architecture.  I've looked at the white paper presented here as well as in UML 2.0 In A Nutshell and I can't figure out the following.

Are the various elements represented in views supposed to be contained in the view packages?

This seems to muddy the waters between what I'd consider a "model" consisting of the elements themselves and the "views" of that model.

Confusion results when one decides to use the same element in diagrams contained in different views.  Why should the element be located in one view rather than another?  As far as I'm aware UML does not strictly specify what can go where in all cases.

Even more confusing would be reusing packages of elements in completely different models.  I assume it makes sense to share through source control the package of components representing some subsystem with another completely different model.

It makes sense to me to maintain a separate package of elements that might be called a "model" of that subsystem independent of any of the 4+1 views.  That "model" might contain diagrams but it would not be, for example, a Logical View.

In general, elements should (I would think) be arranged in top-level (with the views) packages arranged according to system, subsystem, and so forth without regard to any view.

Ultimately, I suppose I'm asking about issues related to the physical design of UML packages that I have not found addressed anywhere else.  I suppose I'm also asking about best practices given that UML 2.x does not (as far as I know) specify what one must do.

Does that make sense?  Or would anyone who really knows UML be hopelessly confused by this?