Importance: Normal Subject: MPI_ALLTOALLW correction To: mpi-core@XXXXXXXXXXXXX X-Mailer: Lotus Notes Release 5.0.3 (Intl) 21 March 2000 From: "Richard Treumann" Date: Fri, 8 Dec 2000 12:50:20 -0500 X-MIMETrack: Serialize by Router on D01ML077/01/M/IBM(Release 5.0.5 |October 12, 2000) at 12/08/2000 12:48:07 PM MIME-Version: 1.0 Content-type: text/plain; charset=us-ascii Sender: owner-mpi-core@XXXXXXXXXXXXX Precedence: bulk I agree with Rolf and Bill G. that the binding for this routine is not ideal as written and should be changed as Rolf proposes. This revives the question - How does such a change get made. There have been several rounds of discussion since July 97, on various topics where those who bothered to participate came to an agreement about what the standard means. We cannot know what opinion those who sat out the discussion would have had. Those who did not participate certainly could claim later that any decision by a small subset of the forum on the email reflector in not binding on the actual meaning of the document. More important, there is no effective audit trail for the decisions that have been made informally. The errata has not been updated in ages. Any implementor of MPI who overlooked some decision made by email discussion would rightly argue that by implementing what was in the MPI-2 document and errata, his implementation was in compliance and everybody else's was wrong. I think this is the first time we have considered an API change that would break existing implementations and applications. Though this change is the right thing to do, I do not think is can be done with the same informality as the earlier resolutions. Is there someone responsible for the mpi-forum web site who could collect the reflector decisions together in an unofficial errata which all of us could review and accept or debate? That would not be a complete solution but would be a big improvement. Since it does not look like there will ever be an MPI 2.1 document I think we need a fall back plan for "corrections" and chapter authors should play a major role if possible. Dick Treumann RS/6000 SP Development IBM Poughkeepsie Unix Development Lab Dept 0lva / MS P963 -- 2455 South Road -- Poughkeepsie, NY 12601 Tele (845) 433-7846 Fax (845) 433-8363