ARCast - Reality Check with the modeling panel

Sign in to queue

The Discussion

  • User profile image
    twmurph
    I had to go back and listen to parts of this several times.  While I enjoyed the topic and the points made by the panelists, it seemed like Jack was a little confused about who he was responding to.

    Maybe I misunderstood what Jack was saying, but to me it sounded like he was diagreeing with what Steve said but crediting Martin with saying it.  Someone please correct me if I heard this wrong.

    Tim Murphy
  • User profile image
    stevek_mcc
    I guess Jack is the only one who can tell us what he was really thinking, but my impression was that he was trying to answer Martin, not me. In particular, Martin's comments that MDA and Software Factories were not diametrically opposed, and that UML could even be seen as a DSL for O-O code, seem to have triggered a fairly vehement denial.

    Ron Jacobs, the host of the podcast, gave Jack and Martin the chance to sort out their differences in a 33 minute podcast conference call podcast just for the two of them. Judging by the abstract there, it seems the confusion you mention was really just between Jack and Martin. In the end I think they had to agree to differ.

    My take would be more aligned with Martin's view, although I understand Jack may have a need to distance himself and Microsoft from the OMG/IBM camp. MDA is a poor form of MDD, with many holes in its story as yet, and Software Factories pulls in a lot of stuff that is basically good programming practice regardless of models, but at heart they're both forms of MDD. UML is a poor example of a DSL, since it focuses on a broad technical implementation domain rather than a narrow problem domain, but it still is a modelling language designed for a particular domain.

    I also disagree with Jack that talking about generating full applications from models is a pipe dream. Maybe he hasn't done it or seen it, but I've seen generating full applications from models work well in many cases over 10 years. In each case, the requirement was a domain-specific modeling language and generators tailored to a single product family of a single company. That's the key, and it can, has and is being done: see some cases here.

    - Steven Kelly, CTO, MetaCase
  • User profile image
    twmurph
    Steve,

    Thanks for responding.  As I said, that was just my take from listening to the show.  I will download the new show and have and see how things unfold.

    I would have to put myself on the skeptical side as well when it comes to generating full applications.  As you say, it could just be that I haven't seen it.

    Thanks again,
    Tim Murphy
  • User profile image
    twmurph
    Steven,

    I have one follow up comment.  I have now listened to the new show and I would have to say that I didn't hear any disagreement between Jack and Martin.  If anything it sounded like, even if they didn't start on the same page, they have a common understanding.

    I think it would be interesting to have you and Jack get on with Ron in the same format.  I think this was much more productive than the voice mail style panel that started the discussion.

    Tim Murphy

Add Your 2 Cents