[Uml-devel] XMI is not standar?

[Andrew Sutton]

On Tuesday 09 September 2003 04:55 pm, Marcos Dione wrote:
>     I'm looking for an app to do some UML design, and I had seen
> poseidon, argouml and now I'm heading to umbrello. but reading you FAQ
> page, I find that XMI is not standardly adopted, or the standard is not
> to-the-bone followed. can you explain more?
>
>     also, I would like to know if arbitrary stereotypes can be used and
> if stereotypes can be 'applied' to methods, attributes and relations.
> I aim to use any UML tool with AndroMDA to generate j2ee code.
>
>     please cc me, as I'm not subscribed to the list.

xmi IS a standard. its not even very complicated. its just kind of hard to get 
standards compliance with it. most uml implementations rely on private data 
or extensions so its easy to break standards compliance if your tools aren't 
written very well...

for example, umbrello's version of xmi store's private diagram information 
within the same xml tree as the actual model (the data for all the classes, 
attributes, packages, etc). argoUML might store its diagrams in SVG (it 
doesn't, but lets pretend). at best, you can have each application read in 
the model data, but it won't be able to re-create the diagrams. at worst, the 
reading won't work because it can't handle the diagram information.

as for your second question: yes. stereotypes can technically be applied to 
just about anything in uml. packages, classes, interfaces, attributes, 
methods, relationships, use cases, states, events, components - you get the 
picture. moreover, you should actually be able to support a set of 
stereotypes for an object (more than one).

for example, consider the following c++ method of some class:

virtual void size() const;

you could give it three stereotypes (assuming the software supported it):
<<c++_operation>>
<<virtual>>
<<const>>

andy