Re: Users of the BON notation among Ada users ?

[Matteo Bordin <matteo.bordin@gmail.com>]

On Jan 13, 8:50 am, Jean-Pierre Rosen <ro...@adalog.fr> wrote:
> Martyn Pike a écrit :
>
> > I have often wondered why HOOD and HRT-HOOD have not remained popular
> > within the Ada community.  Or perhaps they have - anyone care to comment ?

> Because, unlike UML, they are not simply notations, but true design
> *methods*. They drive the design process through a precise engineering
> methodology. This "hampers the creativity" of those who want to write
> software as they please, and not as they are told to.
>
> In the old times, people were proud of having obscure code with a
> comment saying "this is impossible to understand, but don't worry, it
> works". Nowadays, people are proud of covering the walls of their office
> with the most complicated class diagram possible. HOOD is intended to
> limit the complexity one has to deal with. How can you prove you are the
> big guy with such a method?
>
> Hmmm, I think I could go on like this forever ;-)

Because, as a metter of fact, UML is the only standardized 4GL. Or
because back in the days of HOOD, model-driven engineering was
basically moving boxes around and generate skeletons. When it became
evident that you wanted to do more with your models (i.e. analysis,
model simulation, model coverage, etc.) and that the maintenance of
such tools would have required wide use to be economically feasible,
people started looking for tools which could evolve and be supported
over time (or whose language was taught at Universities etc.).

In addition, (HRT-)HOOD, in my opinion lacked a fully specific
execution semantics which could trascend from the way code is
generated - something that other languages like SCADE have and that
UML (in its several flavours like SysML/MARTE) desperately tries to
have without (always) succeding. For the same reason I dislike Ada
profiles as the semantics of a model should be given by the modeling
language itself (the metamodel) rather than by knowing in advance how
you generate code. Then we can discuss which programming language can
better preserve the model semantics at lower abstraction levels - and
in this case Ada is clearly better than C or Java on several aspects.

For what regards enforcing a precise methodology through UML, just
enforce a modeling standard (exactly like you enforce a coding
standard), and you are done.

As a side note, UML reprised (sadly several years later) some good
ideas of HOOD, for example hiearchical decomposition and delegation,
making de facto feasible to design of HOOD-like models in UML (Artisan
indeed sponsored this to let its tools be used for Eurofighter).


> --
> ---------------------------------------------------------
>             J-P. Rosen (ro...@adalog.fr)
> Visit Adalog's web site athttp://www.adalog.fr