I think we should put off thinking about modeling Mastermind in
its own modeling language for the time being. The reasons? First of all
we don't know what kind of functionality or user tasks we are going to have
yet in the design environment. We're shaping the design process as we go so
we're dealing with a moving target here. Second, no one right now can prove
that the models are correct, complete, and are expressive enough to create
fancy interfaces. The proof will come with time and more use of the language.
We should not plan to model the design environment, hence end up limiting the
functionality of the environment due to the limitation of the modeling language.
I think as we build the design environment, we'll see opportunities to save
development time by using the model. That's when we should revisit the issue.
We should try to develop the environment as a traditional program (no model,
that is) to begin with.
If we think we're going to design a perfect system in the first place,
I'm afraid we're going to fail because we will try to manage all complications
at once and it will be out of hand.
--Noi