[Uml-devel] Thoughts about U2 "tree view"

[Andrew Sutton]

On Monday 12 May 2003 02:10 pm, Sebastian Stein wrote:
> Andrew Sutton <asutton at mcs.kent.edu> [030512 20:05]:
> > in my thinking, each metamodel is going to have to define its own actions
> > - it sounds kind of crappy, but there's too much metamodel specific stuff
> > going on. also, we need to constrain which actions are available to which
> > objects.  for example, we shouldn't be able to create attributes within
> > packages or classes within operations (that would be some wild code :)
>
> Why each metamodel? I thought UML and CWM are both defined in MOF, so why
> not just a layer for MOF?

defined in: yes. implemented by: no. we really can't reliably generalize 
semantics between metamodels. we can't say in such in general way, if this 
object is an instance of Model::Class then it has UML::Attribute objects 
within it - because it probably doesn't. for example, UML::Binding is an 
instance of MOF::Class, but can't contain UML::Attributes. these actions are 
going to have to be built up by hand. the good thing is that we can do it 
slowly and piecewise - we don't have to have all the actions defined up 
front.

> I think KAction is good for this kind of stuff. We should at least use
> KAction in our inheritance tree.

agreed. we already for some things :) just not the stuff you're talking about 
quite yet.

andy