[Uml-devel] CASE tools and IDEs

[Andrew Sutton]

> Stereotypes, tagged values and constraints (OCL), yes. But those
> stereotypes are defined in accordance to the UML metamodel, giving them
> semantics.
>
> A tool based on the UML metamodel, making use of profiles can interprete
> those constraints, e.g. prohibit you from adding a "supports" association
> in a CCM Component Interface that inherits another Component.
>
> Without those semantics, you can't validate the model (or, only against the
> UML semantics).
>
> Simply adding a stereotype Foo to a class makes it look like "Foo" but
> still has the same semantics as a normal class.

actually, looking through the UML standard and its interchange (XMI from which 
we plan to generate the UML implementation), there are no actual constraints 
defined for UML. it's actually kind of weird. the standard defines 
well-formedness syntax, but that's not quite the same as constraints - they 
don't have to be followed and they don't get validated.

for UML, at least, the semantics of something is only what we say they are. we 
don't really have a way to back it up. interesting... makes me wonder how we 
can implement the well-formedness rules without jamming code into generated 
interfaces... hmm...

andy