I think it is a bad idea to introduce a new programming abstraction
(sockets) into the MPI spec. (What SETSOCKOPTs are supposed to have
been done to this socket? How does a Fortran program set up this socket?)
In my opinion, more problems are introduced than solved by putting this
C-and-Unix-oriented concept into MPI.
| One observation about the current port_open/accept/connect
| section is that we don't have any good examples of how this
| might be implemented as anything other than thin wrappers
| on the corresponding socket calls.
a) It could definitely be implemented without sockets on current and
near-future shared-memory environments, and even on switch-oriented
ones, with some effort.
b) Even in the cases where it *is* a thin layer over sockets, the advantage
of having this layer is quite positive. Application people I talk to
like MPI partly because it has uniform documentation, a Fortran interface,
and is written about. It transfers some topics from the realm of systems
programming to that of application programming.
My own opinion is that we should leave this thoroughly-discussed-and-voted-on
topic as is, and focus on finishing what is still in flux.