I have read some simple MDA articles, plus their own modeling experience, I feel that there are a few words to say.
First, MDA should be a very good design implementation. On the basis of UML now, I personally think that there are two directions special attention:
First, how to process the front interface interface can be displayed during modeling.
For the C / S system, this issue may not be very serious. However, there is still an interface part that cannot be implemented in modeling. How to deal with this problem? There is a lot of consideration in the future information of MDA and UML. In my development process, it has to be listed separately from the interface design, making it an indispensable key step outside the modeling process.
For the B / S system, this phenomenon is more serious. Because the B / S system itself does not have the concept of interface classes, the interface is just a display location for processing startup information and accepts background feedback information, which does not have a directly related to the processing of related operation control. The interface has to give up in the design process.
Second, the design model to the export process of the code needs to be improved.
It is not possible to achieve the generalized code export. Now there is a kind of reaction to the tool in Rose and related model development, but in fact these reaction work is very wasteful. This requires that the coder has to let go of the work waiting for designers to complete the design modification and then re-export the formation code. This process will seriously increase the amount of model development, causing that everyone is not willing to continue maintenance using design models in the later stage, and modifications to the emotes.
If the MDA or UML tool solves the above two questions, I will be willing to convert the past.
I don't know how everyone thinks?