[Modeling] Class Diagram Specification

James Odell email@jamesodell.com
Sun, 08 Jun 2003 14:06:37 -0400


On 6/3/03 1:03 PM, Wagner, G.R. scribed:

>> When you speak about stereotypes for beliefs,
>> and objects and containers, you are too far involved in the design,
> 
> No, that's not design, it's domain analysis (UML class diagrams can
> be used for domain modeling, for design modeling and for implementation
> modeling). In the domain, there are certain entities (beliefs about
> objects and agents), which I can model as <<belief>>-stereotyped classes.

Yes, I too think of beliefs during analysis;  I.e., beliefs can be described
independently of implementation.


> UML class models is not just another formal language like "Object Z".
> Apparently, you do not understand it's fundamental nature as a
> conceptual modeling language, which extends ER diagrams.

When I work with businesses to develop agent-based systems (and OO systems),
I use class models conceptually, as well.  Here, I literally mean using
classes as concepts.  This is useful for both communication and
specification with humans starting (typically) from the age of seven years.
In this sense, it is similar in spirit to ER diagrams.  (I only distance
agent (and OO) class diagrams from ER diagrams because ER diagrams have a
database and 5th normal form legacy that I do not think useful or
appropriate.)
Having said all that, there are other that prefer other representations.
For example, many academics prefer to communicate and represent their
concepts using Z or OCL or ML. UML class diagrams do not have all the
expressive power of these more-academic languages.  But, its graphical
nature is very useful for the minds of most humans.  So, wouldn't that still
make Class Diagrams "just another formal language"?


Cheers,

Jim