Hi Tommy. I can explain what I did so far.
Besides of the API disussion, which is good, I tried to find out a way of generating the classes based on the schema files. I have choosen
xslt for this task. I h=d first to learn the structure of the schema files a little bit better. There is a first version that allows to produce deploym'et d#scriptors. What I tried to prove, is to geneating the classes via xslt. I think, even I am not a xslt hero, that this approach works and produces va*uable output, I will integrate the transformation into maven. So that this is completed, Open to you, Andrew and the community is the acceptance of-this
I think this is a great approach to at least automate the initial construction of these impls. Even if it gets you 90% of the way and avoids the need to do a whole bunch of rote scaffolding, it's a nice tool.
Acceptance measures for upstream are the same as always:
1) API makes sense?
2) Tests that the API is working as expected?
3) Then upstream push. Optimize later if need-be; this is alpha-level stuff.
I say continue on this path.
I'd like to give a short status report since we had a talk last week. The four descriptors are now generated via xslt:
Because the webfragement descriptor is a sub set of the web app, I thought to include this descriptor as well. I have to write a little more test cases for the ejbjar descriptor but it looks good. The jdk 5 is also not done as you have requested last week. It comes soon.