From: Rolf Hempel <hempel@XXXXXXXXXXXXXXXXXXXXXXXXXXX>
Date: Wed, 10 May 2000 16:37:20 +0200 (MESZ)
To: mpi-core@XXXXXXXXXXX
Subject: Re: Another clarification
Cc: traff@XXXXXXXXXXXXXXXXXXXXXXXXXXX,hempel@XXXXXXXXXXXXXXXXXXXXXXXXXXX
Reply-To: hempel@XXXXXXXXXXXXXXXXXXXXXXXXXXX
Sender: owner-mpi-core@XXXXXXXXXXX
Precedence: bulk

Bill and Dick,

Technically I agree with Dick, i.e. MPI_UNDEFINED would be the right
answer. In practice, however, this would change all implementations
that follow MPICH in this point. While the required change is easy
to do, it might break some user codes which currently work just fine.

In other cases where the standard was not precise enough, the MPICH
interpretation has been accepted as a kind of "de facto standard". I
suggest to do the same here, and therefore second Bill's proposal
(i.e. return zero).

- Rolf
------------------------------------------------------------------------
Rolf Hempel
Manager, C&C Research Laboratories, NEC Europe Ltd.
Rathausallee 10, 53757 Sankt Augustin, Germany
TEL.:  +49 (0) 2241 - 92 52 - 95,    FAX:  +49 (0) 2241 - 92 52 - 99
EMAIL: hempel@ccrl-nece.technopark.gmd.de
WWW:   http://www.ccrl-nece.technopark.gmd.de/~hempel/