BPMN Simplification is a holy grail in some circles, so I couldn't resist posting and linking to Marco Brambilla's blog on the topic here.
In practice, we often use much less of the BPMN spec than is really there.? Someone could probably prove what subset of the spec is still Turing complete for example.? But the challenge as a builder of a BPM engine, is how much of the spec do you need to implement to solve most BPM-solvable problems?? That's? a tricky one to answer correctly in advance.
Check out Marco's slideshow on his blog for more!