Introduction and misc ideas

Cyril Hansen shinjisfriend@netcourrier.com
Wed Jun 5 13:46:02 2002


>
>
>VHDL, ADA, Maude and other specification languages can treat external
>libraries as black boxes, with the interface as glue code. Aditionally,
>Maude and BOBJ can then use the interface specification to prove
>properties about the library, good for proving security issues,
>parallelism etc.
>  
>
I looked at Maude and Project Tatami. These project are possibly the 
"right way" to
fulfill our requirements - I just don't have the necessary background to 
judge them -

I don't know much in the domain of therorem prover (like INRIA's coq) - 
but my impression
was that these systems are still weak and limited and not ready for 
industrial use - but I may be wrong. Additionnaly a strong, formal 
mathemathical culture is required to work with them - wich contradicts 
my idea of a simple unified system accessible to a large number of 
developpers.

>>>I would love to see them implemented as a library for a
>>>widely-used language
>>>
The project could not limit itself to a library - the developper 
experience will be crucial  and we will probably need to start early 
with a modeling/development environment. This environment should be able 
to manage models of applications in constructions and models of 
available APIs, tools and ressources. There is no reason the environment 
should restrain the developper to one language, the design should allow 
the largest freedom.

But at one point the environment itself need to be bootstrapped: A 
language will have to be choosen for this bootstrap. This question may 
become hided completely If we find a good existing environment to start 
from, like Netbeans, ArgoUML, etc..

>But  I think  it would  be better  to try  to be  language independant
>early.
>
>What  about having something  like Objecteering  of Softeam  where you
>model  the  programs in  a  high level  language,  and  have the  tool
>generate the code, the documentation,  whatever, for you.  One can add
>his own methods to generate code or modify the models automatically.


I had a limited experience with previous versions of Together/J and Rational Rose.
These products may have made progress but at that time (99') I had been disappointed by
the level of functionnaly and (lack of) usability. IHMO these tools are probably useful
to manage the complexity of large project in rich companies but they don't make sense when ressources are limited. And technically, round trip engineering basically does not work. How is it going these days... ?

And more importantly, they are not free. In the Free world, there is ArgoUML. It is a very interesting project, but the same points applies to it.

Regards,

Cyril Hansen