We could no longer use text input ?¬?elds to show the time, nor buttons that
the user could click with a mouse. The main application class would no longer
be an Applet, but a Midlet, which worked rather differently. Compilation would no
longer be as simple as ???javac *.java,??? but would require a number of new phases
and con?¬?guration ?¬?les.
Perhaps most challenging, we wanted the same models to be able to generate an
application for either platform. Or rather even more: we did not just want to be able to
get to a state where one set of models could generate both, we wanted the existing
models, built before any idea of MIDP was considered, to work on both platforms. In a
real commercial case, this would have saved the modelers (of whom there would be
many) having to rework each of their models. While achieving this may require more
work in the DSM solution (the metamodel, generators, and framework), the bene?¬?t
would be multiplied over many modelers, each with many models.
Of course, if the claims of DSM are true, and if theWatch DSM solution had been
made according to good DSM principles, this requirement should be feasible. The
watch models would contain only information about what the applications should do,
and no implementation details of how they would do it or on what kind of platform.
Pages:
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593