Hi I've recently received an inquiry regarding this outdated thread,
Given the focus I'm currently turning on my development, see my googlecode page: http://cognescent.googlecode.com
(there is not really much there by the way) you'll see I'm trying to build an underlying framework for BI (Business Intelligence) platforms. My goals are to be able to receive data from any data source, translate it to triples (as much as I can see, any kind of data source is feasible to be translated into triples: relational, NoSQL, XML, etc). Then, given that input triples, infer as much meta data as possible, for example, infer types/classes without any schema information and then build a meta model with this translate-able model to other output models or endpoints.
The idea of MDAfy is still fascinating. Maybe, I'm not sure, this could work with Drools in this way: My model has a set of predefined entities which represents semantic / semiotic relationship. Having a set of 'predefined' rules in Drools for each kind of entity for which the meta model provides 'executable' instances, for me sounds like some kind of meta programming that could be done. Drools could be the engine and the meta model and input data could be kind of script. I'm really not sure if something like this could be done, I'm just thinking out loud.