Hello Andreas
I took a look into the existing perl code and I think the project got more clear now, I will try to describe it in a few simple words:
The current the blend-gen-control does an apt-get update and creates a package pool, from the apt-cache dump output, with all the available packages , then it loads the tasks files and finally generates a control file according to the available packages(it verifies if the the described dependencies in the tasks files exist in the package pool).
The new python design should create a package pool using information from UDD(get available packages' information similar to apt-cache dump's output). In that case it can create architecture dependent package pool(query packages' information for a specific architecture instead of using the architecture from where the blends-dev was called ) and so create architecture dependent metapackages.
Is this approach correct?
Kind regards
Emmanouil Kiagias