Jon,
I think we will more likely encapsulate the plpgsql, so that its not required to write new plpgsql code for normal operations on an object. That is, we'll provide an API close to what exists for the content repository.If you need to add addtional capabilities to an object you can add them in the Tcl layer.
Unless you have some serious performance issue that requires all your operations to be done in the database, coding your specific application requirements in tcl will be much easier.
Regarding XoTcl, we really need some sort of guidelines at to what XoTcl code will look like, and what the best practices to improve code organization with XoTcl are. We aren't there yet.
One problem is the most likely candidates to be "improved" with XoTcl, the pseudo-object apis like template::form etc. are quite stable, and so pervasive in the toolkit, there is a risk to trying to improve it. Maybe someone with more experience can explain the advantage to making these type of changes.
