| This conversation has (appropriately) moved to the Bugzilla issue. I would recommend that you add yourself to the cc list if you have not done so yet.
Joel
On 17 Oct 2008, at 14:19, Joel Rosi-Schwartz wrote: Rich thanks for your thoughts, very helpful in focusing the conversation. On 17 Oct 2008, at 13:58, Richard Gronback wrote: That’s quite the list of requirements you’ve put in the bug. I’m not sure such a tool exists, although perhaps you can get licenses to use a commercial tool for free? I might know one or two that meet your needs.
That's why I said up front:
I deliberately left the requirements hazy because as soon as specified what I
felt was "minimal" I knew that I would knock all of the free ones out of the
market. Here is my "wish list", we as a team will have to make the decision on
which tradeoffs to make:
I do not expect to get the entire "wish list".
My other questions would be on how you expect to use the UML. Do you really intend to use all diagram types? FWIW, UML2 Tools will soon support Sequence Diagrams, though reverse engineering is not planned. Having used a tool that does a nice job of reverse engineering sequence diagrams from code, how often do you really expect to use this functionality? From my experience, it’s difficult to get meaningful sequence diagrams directly from code, though it certainly can be done (they sure look impressive when generated, at first ;).
UML2 would be fine if it had Sequence diagrams and the ability to reverse an existing code base to build the class model. I do not expect it to do Sequence Diagram creation; my experience is this the tends to produce so much noise that needs to be filtered by hand that the saving are nil or negative. While it would be nice to have automatic generation of class diagrams, that is not really essential. It is also very useful if the modelling tool understand libraries so if you reference f.i. an Eclipse class it knows what is being referenced. Again this is probably too much to ask of a free OS tool. Also, if you’re also planning to be model-driven and use a collection of DSLs, I expect Ecore Tools or the GMF Ecore diagram will be sufficient for those models. If you augment these with some UML2 Tools diagrams for communicating the rest of the system, would this not be sufficient? Without having the whole team bought into using UML and taking advantages of a full-blown UML modeling tool/environment, it might be best to start small.
Yes we will definitely be using the DSL and EFM tooling. This is regardless of the tool we select for generic modelling.
Yes, it is not appropriate for us to be dictatorial and state that "Ye shall model!". On the other hand, I really believe that it is an essential tool in the developers toolkit that should be strongly encouraged and nurtured.
Cheers, Joel
HTH,
Rich
On 10/17/08 4:19 AM, "Joel Rosi-Schwartz" <Joel.Rosi-Schwartz@xxxxxxxxx> wrote:
Certainly there is one amongst us who would be proud to take this on their shoulders :-)
Kidding aside, is the lack of response here that the team in general does not believe that modelling adds value to the project? Possibly that is subject that we should be discussing prior to which modeller to use.
Joel
On 13 Oct 2008, at 15:25, Joel Rosi-Schwartz wrote:
Hi,
I would very much like for the entire ORMF team to embrace modelling as both an aid to analysis and design and for documentation. The issue is what tool do we use. The two obvious choices are the eclipse UML2 Tools and Topcased.
The issues with UML2 are that it does not support reverse engineering nor sequence diagrams. This limits its usefulness as a long term complete solution.
I have tried Topcased <http://topcased.gforge.enseeiht.fr/> UML2 Tools. I find that it has usability issues and was also not able to successfully create class models from the reverse engineered model of a fairly simply one plug-in, 6 package, 12 class project. Halfway into walking through the packages it threw a transformation exception. This was using their latest RCP standalone version on Windows, so it is not local configuration issue.
With Topcased there is also the consideration that it is aimed at the "Critical systems Topcased is a software environment primarily dedicated to the realisation of critical embedded systems including hardware and/or software." So it surely has complexity that ORMF does not have to buy into.
If there are recommendations from the team as to other options that would be great.
What I would like is a team member to pick this up, do whatever research and evaluation is required and report back, so that the team can make a decision.
I have opened a bugzilla issue <https://bugs.eclipse.org/bugs/show_bug.cgi?id=250648> to track this work.
Thanks,
Joel
_______________________________________________
ormf-dev mailing list
ormf-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ormf-dev
_______________________________________________
ormf-dev mailing list
ormf-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ormf-dev
_______________________________________________
ormf-dev mailing list
ormf-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ormf-dev
ormf-dev mailing list
ormf-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ormf-dev
|
