I certainly agree about the model. I'm not clear about the extent to which there is commonality below MPI, in the choice of compilers for instance, or even whether this is a significant factor in approaching "use".