[Mpi-forum] Voting results

Jeff Hammond jhammond at alcf.anl.gov
Wed May 30 20:21:57 CDT 2012

On Wed, May 30, 2012 at 7:57 PM, George Bosilca <bosilca at eecs.utk.edu> wrote:
> In general, the term "simple majority" is biased by our origins. In US it simply means over 1/2 of the valid votes cast (so everything except the absentees). Everywhere else, and this includes some other english speaking countries, a simple majority is the greatest number of parts, in other words the choice that got more votes than any other.

Indeed, there are only two countries - none outside of North American
- that use the first definition

> I can hardly believe I'm the only one with this understanding of the simple majority term. Moreover, I remember live discussion where we clearly stated that there __are__ differences between the "abstain" and the "no"vote, and that one must carefully choose her vote. While these discussions were not in the context of first of second voting, they were at least in the context of readings (which is similarly important as it decide the original fate of a ticket). Maintaining several possible interpretations, and changing the counting before a remote vote doesn't make the process more transparent.

I agree completely.  If abstain=no then there is no point in having both.

> Another valid issue raised during this meeting was about the blessing from the user community. Interestingly enough for a standard targeted toward a user community, such an issue was never brought forward. I think users should be an important factor in our decisions, as important as their contributions to the ongoing discussions. And yes the forum should grant them a significant period of time between the moment when we vote a ticket and the moment when this ticket gets into the standard. Our users will clearly take advantage of such a grace period to thoughtfully evaluate the proposal and raise any issues they might have encountered. We should have done this from day one!

No users asked for C++ bindings to be deprecated and a number
protested quite vigorously.  Many users have asked for active messages
and the Forum is has made no progress on this.  Even something as
simple as MPI_Iwin_create, which was requested by multiple users and
had a very valid use case, was shot down in the RMA working group
because it was too hard for Brian to implement.  The notion that users
are a high priority to the Forum is rather farcical as far as I'm
concerned.  I wish this was not the case.


> On May 31, 2012, at 06:10 , Bronis R. de Supinski wrote:
>> All:
>> Hmm. Quite the controversy. However, the rules as enforced
>> in Japan are consistent with my understanding of what they
>> have always been. More importantly, they are consistent
>> with the wording in the bylaws. Here is what Jeff quoted:
>>    a simple majority is defined as a simple majority
>>    of those present and eligible to vote.
>> Those who abstained were present and eligible to vote.
>> They did not vote yes. The effect is that they voted
>> "no" by this definition. If they did not want their
>> vote effectively to be "no" then they should have left
>> the room. I recall several instances in which someone
>> was out of the room (perhaps even momentarily for a bio
>> break) and Jeff recorded their vote as "not present".
>> See the definition above -- they then do not count as
>> present so they do not figure into the required "yes" count.
>> As I stated, my understanding of the rules is consistent
>> with the interpretation used in Japan. I would object to
>> any other interpretation since the by-laws are actually
>> clear on this point. I agree that the by-laws should be
>> clear in general; while I think they are clear, I would
>> not object to a clarifying statement being added to the
>> effect that "abstentions are effectively negative votes."
>> I think we have many other issues that should be made
>> concrete in the by-laws and this is the least important.
>> What is required to pass a first reading is probably the
>> most obvious issue.
>> Bronis
>> On Wed, 30 May 2012, Fab Tillier wrote:
>>> Jeff Squyres wrote on Wed, 30 May 2012 at 12:12:26
>>>> On May 30, 2012, at 2:57 PM, Jeff Hammond wrote:
>>>>>> The fact that some votes were still recorded as 'abstain' is an indication
>>>>>> that this bylaw change was half baked.
>>>>> Especially when the meeting is attended by so few people due to the
>>>>> location.  It seems like a weasel tactic to pick a remote location to
>>>>> change the by-laws with a single vote.
>>>> To be clear, the process document states:
>>>>    For the purposes of voting, a simple majority is defined as a simple
>>>>    majority of those present and eligible to vote.
>>>> In the context of the document, the phrase "simple majority" is used to
>>>> describe what is needed for ballots to pass; this sentence is attempting to
>>>> define that phrase.  So even though the above sentence looks like a circular
>>>> definition, I think it's really an open-ended definition (e.g., a google search
>>>> for "simple majority definition" turns up both definitions).
>>>> I was not there and don't know *exactly* what happened, so I'll refrain from
>>>> commenting further.
>>> If the bylaws are vague, we should clarify them.  We should not however reinterpret them at each meeting, and should all agree on a proper interpretation and stick to it, such that ambiguity is removed going forward.  Allowing our bylaws to be vague enough to afford a re-interpretation at each meeting does nobody any good.
>>> -Fab
>>> _______________________________________________
>>> mpi-forum mailing list
>>> mpi-forum at lists.mpi-forum.org
>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
>> _______________________________________________
>> mpi-forum mailing list
>> mpi-forum at lists.mpi-forum.org
>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
> _______________________________________________
> mpi-forum mailing list
> mpi-forum at lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum

Jeff Hammond
Argonne Leadership Computing Facility
University of Chicago Computation Institute
jhammond at alcf.anl.gov / (630) 252-5381

More information about the mpi-forum mailing list