On Wed, Nov 29, 2000 at 10:30:05PM -0500, Raul Miller wrote:
> On Thu, Nov 30, 2000 at 11:16:14AM +1000, Anthony Towns wrote:
> > If you still require N initial votes and 1 final vote, it has no benefit
> > over the current wording, at all, since supermajorities only apply to
> > final votes, which are forced to have simply Yes/No/Further Discussion
> > as their options.
> This is an allowed option, but it's not required.
Gag. Can you *please* try to follow?
A clear and obvious interpretation of the constitution is that it is
possible to decide on an issue by having N + 1 votes, initially N votes
to decide on what form the resolution should take, followed by 1 vote
to ensure quorum and supermajority requirements are met, and to see if
the resolution is passed.
Given a particular sentiment in the Debian community with regard to these
possibilities, this procedure results in a particular outcome.
If, instead, for expediency perhaps, you choose the method you describe
(a single vote with all the options, supermajorities and quorums all taken
into account as you describe) you can end up with a *different* result.
That is, supporters of, say, option A have a reason to argue for the N+1
votes to take place, whereas supports of option B have a reason to argue
for the resolution to be decided by a single vote, since whichever way the
vote is processed can dictate who wins.
> However, it could also be handled in a different fashion. This
> would depend on the way the ballots are prepared.
Yeah, sure: being able to handle to vote differently (and thus get it
over and done with quicker) is a good thing, but not if it changes the
result of the vote.
The decision should be made based *solely* on the wishes of the community,
not based on whether the secretary feels like being expedient or not.
> > If you'd rather that I'd phrased it as "you instead just have a final
> > vote, with the pairwise preferences", you're welcome to take it as that.
> "It?" [Sorry, there's too many concepts flying around for me to parse
> this reasonably.]
How, exactly, do you propose the non-free vote, say, be resolved with a
single vote, then?
> > Again, the three things that need to be changed in the constitution,
> > IMO, are:
> > a) resolving circular ties needs to be decided on and spelled out
> I think I dealt with that.
Yes.
> > b) votes with multiple options need to be able to be handled by a
> > single vote
> That's already possible.
Only by secretarial interpretation.
> > c) supermajority requirements need to be updated to cope with (b)
> There was a question about exactly what supermajority meant in this
> context, and while I recognize the interpretations which give rise to
> that ambiguity, I think I see what is supposed to happen there. I like
> to believe that my alternate phrasing is unambiguous about this.
It's unambiguous, but it has undesirable properties (ie, as well as voting
in a particular way, supporters of one side or the other have a stake in
how the issue is voted on also, since one way makes one group more likely
to win, and another way makes another group more likely to win).
To be specific: supposing no one actually minds getting rid of non-free,
everyone wants to see it moved off the mirrors, in general, but there is a
reasonable minority of people who would rather see it kept on some out of
the way machine, just for old times sake.
In the N+1 vote system, the votes go like:
60 Remove, Move, Further Discussion
40 Move, Remove, Further Discussion
Remove wins, and the final vote is:
100 Yes, Further Discussion, No
and Remove wins.
If, however, some of the 40 people connive together privately, and
convince the secretary to decide the issue in a single vote, using
your interpretation of the supermajority clause, they vote instead goes
something like:
60 Remove, Move, Further Discussion, Do Nothing
40 Move, Remove, Further Discussion, Do Nothing
and the results are scaled to:
Remove dominates Further Discussion & Do Nothing, 33.3 to 0
Move dominates Further Discussion & Do Nothing, 100 to 0
Move dominates Remove, 40 to 20
and, the 40 people who prefered Move rejoice having won the vote.
Surely you agree that a minority of people being able to subvert the
resolution procedure to get what they want instead of what the majority
want is a bad thing?
> On Thu, Nov 30, 2000 at 11:41:26AM +1000, Anthony Towns wrote:
> > One problem if you don't have further discussion win more often than it
> > perhaps should is as follows:
> > Suppose you have three options on your ballot, A, B and F. A requires a
> > 3:1 supermajority. Sincere preferences are:
> > 60 people order the options A, B, F
> > 40 people order the options B, A, F
> > in which case A would win by dominating B 60:40 and F 100:0 (33.3:0).
> Since a vote for B is a vote against A, I completely disagree with
> this assesment.
A vote for B is *NOT* a vote against A.
A vote for B over A says "I would prefer it if Debian resolved <B>",
not "I think it would be unacceptable if Debian resolved <A>".
> After substituting meta-syntactic variables, this would read:
"Yes" is not a metasyntactic variable, however. (If it were, section
A.3(2) would have to be read as requiring the final vote to have three
options, and that supermajority requirements only apply to the first
option).
> [You might try to claim that some of these are not meta-syntactic
> variables -- but that would be equivalent to the claim that there is no
> 3:1 supermajority requirement for A, in this vote.]
There are three ways of handling supermajorities under discussion, afaict:
* Two (N+1) votes, the latter being Y/N/F with Y requiring the
supermajority, and no supermajority requirement in the former
vote.
* A single vote, where the pairwise preferences for A against
"Further Discussion" (only) are scaled according to A's
supermajority requirements.
* A single vote, where the pairwise preferences for A against
all other options are scaled according to A's supermajority
requirement. [0]
The first two of these methods can be made to have the same results in
all cases, given a particular sentiment among the electors. The latter
method will give different results in a number of cases.
Cheers,
aj
[0] I'm not sure your proposal was made entirely clear how you would
scale the pairwise preferences for A versus B, when A requires a
3:1 supermajority and B requires a 2:1 supermajority. It doesn't
affect any of the discussion so far, nor does it resolve the problem
identified above.
--
Anthony Towns <aj@humbug.org.au> <http://azure.humbug.org.au/~aj/>
I don't speak for anyone save myself. GPG signed mail preferred.
``Thanks to all avid pokers out there''
-- linux.conf.au, 17-20 January 2001
Attachment:
pgpKi8yxVQCpm.pgp
Description: PGP signature