AW: [Objectbridge-jdo-dev] Intermediate results

[Budde, F. <fri...@vi...>]

Hi Thomas,
> that's great! I loaded your model with the just released 
> Poseidon 1.1 CE
> without problems.
> I intend to have more and more OJB details modeled in 
> Argo/Poseidon CE.
So do I :-)

> 
> I will setup a CVS soon to have this model shared. 
Very good!

> I suggest to keep not everything in one model but to have 
> (say) one .zargo
> per package.
> The reason: .zargo can only be handled as a monolithic binary 
> file by CVS.
> Thus concurrent edits are very difficult to maintain. What do 
> you think?
That's a difficult question. Since there are a lot of inter-package
dependencies I'd suggest to build a single model (at present). 
If we could get rid of some of these dependencies and if we
could define some (kind of) components with one or at most a few facade(s)
which
control access to the component we'd be able
to split the model into separate models for each component.

> I've a question how to use Poseidon diagrams with html (or xml)
> documentation: Until now I simply exported  GIF or PNG files 
> that have to be
> kept up to date manually :-(
> Is there a way to do this automatically or to have some kind 
> of hyperlink
> mechanism from the HTML document into the Poseidon model?
None that I know of. I do it the same way as you do it :-(. 
A kind of scripting language for Poseidon would be nice.

> Another question: how to keep the model in sync with the 
> code? I know Argo
> supports reverse engineering, but does it help to show diffs? 
> 
Actually I haven't used reverse engineering and I'd propose to 
keep on doing it that way. Here is my approach to modeling OJB
- Only model important (logical) features. Keep technical 
  details in the javadoc.
- Use round-trip engineering only for single classes or
  packages and remove or hide technical details after
  importing the class(es).
- Keeping model and code in sync has to be done (mostly) manually.
  It shouldn't be too difficult if the model doesn't contain
  to many technical details (s.above) and if we can find out
  what changed by comparing different cvs versions of the code.

  Frido