Hi all, Attached is a new version of the rules document, which includes the feedback so far (changes of changes are in blue). This new version does allow for a shortcut (but under very severe limits) and also always requires two votes. Let us know what you think. As mentioned before, we would like publish the final version with all feedback by May 4th - within four weeks of the forum - with the goal to pass it in June. So, if there are any concerns or comments, please let us (Jeff and Martin) know. Also, we’ll hold a webex discussion on these rules on April 27th at 8am PDT. Here is the dial-in information: Webex link: https://cisco.webex.com/ciscosales/j.php?MTID=mc87e1ef49d7e73dd9e22ae94a624 dba2 Webex password: mpi Anyone interested is welcome to join, Martin ________________________________________________________________________ Martin Schulz, [email protected], http://scalability.llnl.gov/ CASC @ Lawrence Livermore National Laboratory, Livermore, USA On 4/14/15, 10:17 AM, "Steven Oyanagi" <[email protected]> wrote:
Hi,
After some thought, I think the Forum should always have two votes for the final draft of the standard. As Martin previously pointed out, the ratification process is important enough to warrant the two vote rule similar to regular text changes. Two votes would allow more members to vote on the final draft. As long as they don’t miss two meetings in a row they would get to vote on some form of the final draft. A number of member institutions don’t go or rarely go to the international meeting so they could potentially miss a single final ratification vote. If I recall correctly, the final vote for the MPI 3.0 standard was scheduled for meeting in Europe, though for some reason that I don’t recall the final vote occurred by e-mail.
If the Forum does decide to have a “fast path” approval process, a vote should be required to use the fast path and it should not be at the same meeting as the single standard ratification vote. - Steve
-----Original Message----- From: Rolf Rabenseifner <[email protected]> Reply-To: Main MPI Forum mailing list <[email protected]> Date: Tuesday, April 14, 2015 at 11:44 AM To: Main MPI Forum mailing list <[email protected]> Subject: Re: [Mpi-forum] Proposed Update the MPI Forum Rules
If the changes would have been smaller, then our March meeting would have finalized MPI-3.1 - I'm pretty sure.
Yes, I also feel that there should be a short path. For this, the voting slots for RCM may be on the 3rd day, but should be movable to the last day and substitute the FRM voting slot.
Rolf
----- Original Message -----
From: "schulzm" <[email protected]> To: "Main MPI Forum mailing list" <[email protected]> Sent: Tuesday, April 14, 2015 6:06:39 PM Subject: Re: [Mpi-forum] Proposed Update the MPI Forum Rules
Hi Aurelien, all,
Jeff and I have discussed this as well, but thought the ratification process is important enough to also warrant the two vote rule, as do regular text items. The reasoning was that there will always be some changes in the RCM (even if they are minor) and this would give people a time to think about them. However, that is certainly a good point for discussion to allow a quick path if there are really no changes pending (something like allowing the RCM and FRM to be at the same meeting with the two votes separated by at least one night). How does the rest of the group feel about this?
Thanks!
Martin
________________________________________________________________________ Martin Schulz, [email protected], http://scalability.llnl.gov/ CASC @ Lawrence Livermore National Laboratory, Livermore, USA
On 4/14/15, 6:00 AM, "Aurélien Bouteiller" <[email protected]> wrote:
It seems that the new rules make the preparation of the final version always require 2 meetings. We should keep the possibility open for a 1 meeting final version, for the case when all tickets have been voted and implemented long ago and we only need to vote, like it happened for 3.0. One could argue that the last release meeting where we had still semantic patches on our plate is the outlier rather than the norm.
Aurelien
-- Aurélien Bouteiller ~ https://icl.cs.utk.edu/~bouteill/
Le 14 avr. 2015 à 01:26, Schulz Martin <[email protected]> a écrit :
Hi Steve,
Sorry, I should have made this clear - our proposal would be to consider the March meeting a successful RCM and then, if the rules are accepted, hold the Chicago meeting in July as the FRM.
Thanks for pointing this out,
Martin
______________________________________________________________________ _ _ Martin Schulz, [email protected], http://scalability.llnl.gov/ CASC @ Lawrence Livermore National Laboratory, Livermore, USA
On 4/13/15, 8:16 AM, "Steven Oyanagi" <[email protected]> wrote:
Hi,
A possibly dumb question, but one that needs clarification for those of us who were not at the March MPI Forum meeting. The new voting rules propose a ³Release Candidate Meeting (RCM)² and a ³Final Ratification Meeting². For MPI-3.1, is the March meeting considered to be the ³Release Candidate² meeting and we are on track to have final ratification of MPI-3.1 in June, or will June be the RCM and final ratification would occur in September? - Steve
-----Original Message----- From: Martin Shulz <[email protected]> Reply-To: Main MPI Forum mailing list <[email protected]> Date: Monday, April 13, 2015 at 1:13 AM To: Main mailing list <[email protected]> Subject: [Mpi-forum] Proposed Update the MPI Forum Rules
Hi all,
As discussed at the last forum meeting, Jeff and I drafted an updated version of the MPI rules/voting document that we want to propose to the MPI forum and that, if accepted, is intended to cover the MPI 3.1 ratification. The document is attached and all changes compared to the previous document are marked in red. The idea was to basically write up the process we followed at the last meeting and that most seemed to like. Questions and comments are, of course, welcome our intent is to publish a final version with comments included by May 4th, i.e., 4 weeks before the June forum, and then put this document up for a vote at the meeting.
Thanks!
Martin
____________________________________________________________________ _ __ _ Martin Schulz, [email protected], http://scalability.llnl.gov/ CASC @ Lawrence Livermore National Laboratory, Livermore, USA
_______________________________________________ mpi-forum mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
_______________________________________________ mpi-forum mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
-- Dr. Rolf Rabenseifner . . . . . . . . . .. email [email protected] High Performance Computing Center (HLRS) . phone ﯯ(0)711/685-65530 University of Stuttgart . . . . . . . . .. fax ﯯ(0)711 / 685-65832 Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307) _______________________________________________ mpi-forum mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum