I agree with you that BPDM is a great meta-model that captures all the
aspects that we will need. But it is much too big when you want to
implement all aspects in a few month. Using BPDM would take us years to
come up with a presentable solution. Therefore, I agree with Marc in one
of his former posts, where he said that we should more focus on existing
tools and how they can be changed in order to meet all requirements for
XPDL, BPEL, etc. than to start from scratch and throw away all existing
code.
The work from Steve is a great starting point and gives us hints what we
will need to change on the current meta-model in order to generate XPDL in
the end.