On Aug 25 2012, Rolf Rabenseifner wrote:
However, forbidding it being done everywhere isn't a major restriction.
Yes, this was decided already in MPI-1.
The trouble about MPI-1 was that, back then, practical use still had to allow for K&R C compilers and Fortran 77 compilers, so some people HAD to declare MPI names just to use them. It was entirely reasonable not to specify restrictions in detail, because they would have been almost indescribable. Anway, what the heck. Real language experts will know what is possible. Everybody else should write conservative code because it is too easy to fall into a trap by accident. In my experience this has never actually been a problem except for programs that were being too clever by half, and those got what they deserved! Regards, Nick Maclaren.