[Elections] A simple SqF board elections solution using SqP

[Peter Crowther]

[NB there is a suggestion for a Web site that may make our lives easier
at the bottom of this message]

> From: [...] Ken Causey
> I'd like to float this as a solution to the quickly 
> approaching vote for the 2006-2007 Squeak Foundation Board.

I note that.  Comments inline.  Once again, apologies for lateness.

> 1.  Solicit for people interested in sitting for the 2006-2007 SqF
> board.  This would be made up of 2005-2006 board members still
> interested in serving and any new prospects.  Exactly how you (the
> Elections team) handle these selection of new prospects is up to you.

We need to decide this pronto.  I propose that we allow anyone who can
get nominated and seconded to stand.  We need three pieces of
information then: the agreement of the candidate, the nomination from
the nominator, and the second from the seconder.  I'd require the
existing Board members to go through the same process, i.e. they are
*not* automatically included in the election.

I'd also suggest that any individual may only nominate or second one
candidate, but that may be rather contentious.

> We specify a SqP rank
> (apprentice or journeyman, this is open to debate) that is 
> eligible for voting.

Use of SqP rank is contentious.  I don't think we have any way of
resolving that contention and achieving consensus.  I also think that
the established mechanism for doing that (kick the decision upstairs) is
inappropriate in this case.

I genuinely have no idea how we proceed past this impasse.

> However I think even if the 'process' is done 
> outside of public
> view that exact data used is public and anyone that cares to make a
> snapshot of the people page at midnight GMT on the 7th can verify the
> data generated.

We'd need to snapshot the 'raw' ratings - who rates who as what - to
ensure transparency, but yes.

>   b.  Each listed member is given a fraction value to designate his or
> her desirability to the voter.  The top listed candiate is 
> given a value
> of 10/10 or 1.  The second listed candidate is given a value of 9/10,
> the third 8/10, the fourth 7/10, and so on until the 10th listed
> candidate is given the value of 1/10.  If the voter does not list all
> candidates then the listed candidates are given their 
> appropriate values
> as above and all other non-listed candidates are assumed to 
> have a value
> of 0.
> 
>   c.  A running count is maintained for each candidate as the 
> votes are
> processed summing the fraction value in each vote for that candidate.
> 
>   d.  The 7 board members are selected by taking the 7 candidates with
> the highest sums.  A comment is posted as quickly as possible on the
> 15th of february with the total sum for each candidate and 
> listing the 7
> new board members.

This is a very uncommon voting mechanism.  Can we use something more
common?

Variants of Condorcet, proposed by Lex and used by Debian, appear to be
appropriate for the election of multiple candidates to posts; we could
use something like http://www.cs.cornell.edu/andru/civs.html to run our
election for us if we could identify our list of voters.

> This mechanism is completely transparent and can be rechecked 
> by anyone at any time to verify the results.

Indeed.  For recording and counting votes, however, I'd settle for a
system where the system's owner has no interest in tampering with the
result.  The Web site referred to above has this property.

		- Peter