On Fri, May 16, 2008 at 1:39 AM, Thomas N. <[EMAIL PROTECTED]> wrote:
>> Is there any reason to keep both?  Is it too risky to just commit to
>> the transition after 0.26 is released and move everything over to the
>> new module?  My preference would be to keep the transition period as
>> short as feasible to minimize the extra work.
>
> I'd suggest a step-by-step transition, based on the moving of parts of the 
> Java support. Java import is the first available part, so something I can 
> definitely promise is:
>
> Step 1: After 0.26 (when working towards 0.27.1), switch to the module for 
> Java sources import, which includes:
> - adding the module to the ArgoUML deployment
> - freeze org/argouml/uml/reveng/java (and later remove it)
> - for Java source import, only work in the module (fixes/enhancements)
> - in the module, add an ant task that only packages the import code into the 
> module (alternative: use svn branching)
>
> Step 2 then would be the move of another part (e.g. code generation) with 
> similar steps, and so on. A roadmap has to be made.
>
> What do you think, is this feasible?

That sounds good to me, but Linus needs to agree since he's the one
who does the packaging of modules for the builds.

I think we can probably do both code generation and reverse
engineering in one step since they are both already separate modules.
Moving them should just be a mechanical process.  Additional steps
like separating out the Java critics will require more work and can be
done in stages.  Of course everything that gets moved is going to lose
its entire CVS/SVN history since they'll be moving to a separate SVN
repository, but there's nothing we can do about that with our
fragmented project structure.

Tom

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to