Re: [mdr-users] UML 2.0 Support

Hi Matthias,

Matthias Bohlen wrote:

>Hello Martin and the others,
>
>Thursday, March 31, 2005, 2:03:06 PM, you wrote:
>
>MM> at this point of time, we can think about creating a compatibility
>MM> on the level of XMI - by creating an XMI reader and writer that
>MM> will be able to read/produce MOF2XMI documents into/from a MOF 1.4
>MM> repository. For full MOF 2.0 support we would need JMI 2.0.
>
>does this mean that it is possible to read a UML 2.0 model (which is
>formatted as XMI 2.1) into MDR which is a MOF 1.4 repository? And then
>access the resulting AST via the old JMI 1.x? This would indeed be
>a good interim solution for us.
>  
>
Theoretically this should be possible. However, the final XMI files for 
UML 2.0 are not available yet as far as I know.

>MM> We haven't started evaluating the first option yet. At this point
>MM> there are not enough use-cases for us to invest in that.
>
>Well, for us it is THE MAIN USE CASE! AndroMDA is a code generator
>that follows the MDA specs, so reading models into an AST and
>generating code from it is most important to us. We have been using
>MDR happily for more than two years now and would like to continue to
>use it.
>
>We only need the following steps:
>
>* read UML 2.0 metamodel and generate JMI interfaces from it
>* read XMI 2.1 and build an AST
>* get access to the AST via JMI interfaces
>* all this for MagicDraw and Poseidon
>  
>
AFAIK, MagicDraw and Poseidon still use UML 1.4 XMI (although e.g. 
MagicDraw claims to support UML 2.0 notation).

>MM> Anyway, it should not be that hard - anyone from the community is
>MM> welcome to contribute. Martin
>
>Martin, which steps would be required to implement the list above in
>MDR? If you give us a "shepherd" person, we could contribute a
>solution.
>  
>
Give it a few weeks. It seems that a work on JMI 2.0 will get started 
very soon. And also we need to get a better idea on what tools are going 
to support UML 2.0 in what form to be able to decide what we should 
support - it does not make sense to build support for something that 
nobody will be able to interoperate with...
Martin