Re: Board of directors membership election

The board of directors of Software in the Public Interest has three
vacancies which need to be filled. At its last meeting the board
accepted resolution 2003-10-14.iwj.6 which regulates an election to
fill those positions and which temporarily reappointed Ian Jackson and
Martin Schulze to fill two vacancies for the next ninety days, or until
elected new members are installed, whichever comes earlier.

The election will be done using the same procedure as used for the
previous election. The voting procedure will be as follows:

* From October 23 to end of November 6 (UTC) contributing members
can declare their candidacy by submitting a position statement to
the SPI secretary at secretary(at)spi-inc(dot)org .

* From November 7 to end of November 21 (UTC) contributing members can
vote on the candidates.

Votes will be counted using the "Condorcet" election method system which
will be used to select the most preferred candidate. Conceptually, the
election will be broken into a a series of pairwise races between each
possible paring of the candidates. If one candidate beats each of the
others in pairwise races, that candidate wins. Otherwise, the
"Cloneproof/Schwartz Sequential Dropping" method is employed to choose
the most preferred candidate from those remaining. Some background
reading on preferential voting and Condorcet is available online at
http://www.electionmethods.org/ .

Wichert.

--
Wichert Akkerman <wichert(at)wiggy(dot)net> It is simple to make things.
http://www.wiggy.net/ It is hard to make things simple.

On Thu, Oct 23, 2003 at 02:26:57PM +0200, Wichert Akkerman wrote:
> Votes will be counted using the "Condorcet" election method system which
> will be used to select the most preferred candidate. Conceptually, the
> election will be broken into a a series of pairwise races between each
> possible paring of the candidates. If one candidate beats each of the
> others in pairwise races, that candidate wins. Otherwise, the
> "Cloneproof/Schwartz Sequential Dropping" method is employed to choose
> the most preferred candidate from those remaining. Some background
> reading on preferential voting and Condorcet is available online at
> http://www.electionmethods.org/ .

Looking over the list of votes last time, I noticed that a number of
people voted in a way that meant their vote was worth little or less
than they probably wanted. The instructions were pretty clear on the
fact that we're using a preferential voting system, that preferences
alone is what matters, and that if you vote for only one candidate and
that's it, your vote means nothing. This is different than the way
that Debian (which also uses Condorcet) handles unranked
choices. Perhaps the instructions weren't as clear as they could have
been.

I'd be happy to further clarify the voting instructions and add a
little bit of *bold face* to minimize the impact of this type of
confusion this time around.

Regards,
Mako

--
Benjamin Mako Hill
mako(at)debian(dot)org
http://mako.yukidoke.org/

On Thu, Oct 23, 2003 at 07:18:34PM -0700, Benj. Mako Hill wrote:
> I'd be happy to further clarify the voting instructions and add a
> little bit of *bold face* to minimise the impact of this type of
> confusion this time around.

This was the problem where if you had 4 options and only voted for AC
then in fact your vote really meant nothing unless it came down to only
A and C?

Why not fix the system so that it makes more sense for humans?
AC means first A then C then B and D equally last. It's what the vast
majority of people would expect. A vote of C means a vote for C only,
with ABD last.

All it takes is to have a rule that states that any unmentioned options
imply all of those options have been put equally last. Having the same
way as last time is just plain confusing and does not give any
advantages. It's a great way of disenfranchising SPI members.

- Craig
--
Craig Small GnuPG:1C1B D893 1418 2AF4 45EE 95CB C76C E5AC 12CA DFA5
Eye-Net Consulting http://www.enc.com.au/ MIEE Debian developer
csmall at : enc.com.au ieee.org debian.org

Previously Craig Small wrote:
> Why not fix the system so that it makes more sense for humans?
> AC means first A then C then B and D equally last. It's what the vast
> majority of people would expect. A vote of C means a vote for C only,
> with ABD last.

We're using condorcet voting, which does not have a concept of equal
ranking except in the no preference sense. People should really just
read the instructions, and the voting system should probably print
a fat warning saying 'your vote has no meaning' if people select
a single option.

Wichert.

--
Wichert Akkerman <wichert(at)wiggy(dot)net> It is simple to make things.
http://www.wiggy.net/ It is hard to make things simple.

wichert(at)wiggy(dot)net (Wichert Akkerman) writes:

> a fat warning saying 'your vote has no meaning' if people select
> a single option.

In real life, if someone marks only one candidate, it's hard to imagine that
their true intent was to express no preference.

Treating such ballots as worthless seems overly pedantic to me.

The Debian solution of treating all unmarked entries as being of equivalent
weight that is less desired than any marked entries seems like a much more
user-friendly approach. How about we update SPI documentation and processes
to do that instead?

Bdale

Previously Bdale Garbee wrote:
> Treating such ballots as worthless seems overly pedantic to me.

Fact remains that you have to with condorcet since it doesn't allow
equal ranking of all the other options.

Wichert.

--
Wichert Akkerman <wichert(at)wiggy(dot)net> It is simple to make things.
http://www.wiggy.net/ It is hard to make things simple.

On Fri, Oct 24, 2003 at 07:53:04PM +0200, Wichert Akkerman wrote:
> Previously Bdale Garbee wrote:
> > Treating such ballots as worthless seems overly pedantic to me.
>
> Fact remains that you have to with condorcet since it doesn't allow
> equal ranking of all the other options.

You just need to add the default option after all the votes expressed,
and then you don't care about the ranking of the other options.

So a choice of A, B, C and D being the default option, a vote of A only,
could be interpreted as ADBC, and a vote of CB as CBDA.

Or did i misunderstand something, and this is not the same condorcet we
overdiscussed on the debian-vote list all this past time ?

Friendly,

Sven Luther

Previously Sven Luther wrote:
> So a choice of A, B, C and D being the default option, a vote of A only,
> could be interpreted as ADBC, and a vote of CB as CBDA.

A vote of A could be ABCD, ABDC, ACDB, ACBD, ADBC or ADCD. All could
lead to a different result.

Wichert.

--
Wichert Akkerman <wichert(at)wiggy(dot)net> It is simple to make things.
http://www.wiggy.net/ It is hard to make things simple.

On Fri, Oct 24, 2003 at 01:44:02PM +0200, Wichert Akkerman wrote:
> We're using condorcet voting, which does not have a concept of equal
> ranking except in the no preference sense.
Do you mean no preference at all?

> People should really just
> read the instructions, and the voting system should probably print
> a fat warning saying 'your vote has no meaning' if people select
> a single option.

A warning at least would be useful. What does Debian use for it's GR?
It appears to allow incomplete ballots.

So SPI uses one sort of voting, which must not have blank options, and
Debian uses another sort of voting, which does.

"In the brackets next to your preferred choice for group name, place a
1. Place a 2 in the brackets next to your next choice. Continue till
you use 3 for your last choice. You may leave choices you consider
unacceptable blank."

You should be able to easily see why the SPI voting system has a problem.

- Craig
--
Craig Small GnuPG:1C1B D893 1418 2AF4 45EE 95CB C76C E5AC 12CA DFA5
Eye-Net Consulting http://www.enc.com.au/ MIEE Debian developer
csmall at : enc.com.au ieee.org debian.org

wichert(at)wiggy(dot)net (Wichert Akkerman) writes:

> Previously Sven Luther wrote:
>> So a choice of A, B, C and D being the default option, a vote of A only,
>> could be interpreted as ADBC, and a vote of CB as CBDA.
>
> A vote of A could be ABCD, ABDC, ACDB, ACBD, ADBC or ADCD. All could
> lead to a different result.

If D is none of the above, then treating A as AD is probably the simplest
solution to avoid discarding votes cast without assuming more than the vote
cast implies?

My concern is that throwing a vote of A away is "unfriendly" even if it is
"correct" in some sense.

Bdale

On Mon, Oct 27, 2003 at 09:04:24AM +1100, Craig Small wrote:
> On Fri, Oct 24, 2003 at 01:44:02PM +0200, Wichert Akkerman wrote:
> > We're using condorcet voting, which does not have a concept of equal
> > ranking except in the no preference sense.
>
> Do you mean no preference at all?

Right. Condorcet works by comparing preferences in bitwise pairs. In a
"pure" Condorcet implementation (AIUI), if you don't list a candidate,
you simply won't be expressing a preference between that candidate and
any others. If you list only one candidate, you haven't expressed
preference at all. If you leave one person off, you simply haven't
affected any of the sub-races between that candidate and any others.

> > People should really just read the instructions, and the voting
> > system should probably print a fat warning saying 'your vote has
> > no meaning' if people select a single option.
>
> A warning at least would be useful.

There was a very clear warning last time. I agree that we should at
least *bold* a line of it or something because evidently, some people
didn't read it.

> What does Debian use for it's GR? It appears to allow incomplete
> ballots.

I'm not an expert on the Debian system but I believe that Debian uses
a modified version of Condorcet that allows for things like quorum,
super majorities, default options and a couple little tweaks to make
sure it works in some corner cases (I believe).

> So SPI uses one sort of voting, which must not have blank options, and
> Debian uses another sort of voting, which does.

The two systems are already very similar.

According to the by-laws, it's up to Wichert. It seems that ordering
all unranked pairs as equally ranked (same as no preference really)
but lower than the last ranked choice makes sense. Personally, I don't
care too much as long as it's documented well, which was my only point
in bringing up this issue in the first place.

There's only so far we should go for people who can't be bothered to
read a half-paragraph of instructions on the page that they enter
their vote. :)

Regards,
Mako

--
Benjamin Mako Hill
mako(at)debian(dot)org
http://mako.yukidoke.org/

On Fri, Oct 24, 2003 at 01:44:02PM +0200, Wichert Akkerman wrote:
> Previously Craig Small wrote:
> > Why not fix the system so that it makes more sense for humans?
> > AC means first A then C then B and D equally last. It's what the vast
> > majority of people would expect. A vote of C means a vote for C only,
> > with ABD last.
> We're using condorcet voting, which does not have a concept of equal
> ranking except in the no preference sense.

It's trivial to deal with this issue in Condorcet voting, though. When
creating your matrix, you only put in values where a preference has been
expressed, eg for a vote [12--] amongst candidates A, B, C and D, you add:

A B C D
( 0 1 1 1 ) A
( 0 0 1 1 ) B
( 0 0 0 0 ) C
( 0 0 0 0 ) D

to your preferences matrix, instead of something like this, for the vote
[1243]:

A B C D
( 0 1 1 1 ) A
( 0 0 1 1 ) B
( 0 0 0 0 ) C
( 0 0 1 0 ) D

The complexity in the Debian scheme about handling circular ties (the
sequental dropping and so forth) is completely independent to this.

Cheers,
aj

--
Anthony Towns <aj(at)humbug(dot)org(dot)au> <http://azure.humbug.org.au/~aj/>
I don't speak for anyone save myself. GPG signed mail preferred.

Australian DMCA (the Digital Agenda Amendments) Under Review!
-- http://azure.humbug.org.au/~aj/blog/copyright/digitalagenda

Wichert Akkerman writes ("Re: Board of directors membership election"):
> Previously Bdale Garbee wrote:
> > Treating such ballots as worthless seems overly pedantic to me.
>
> Fact remains that you have to with condorcet since it doesn't allow
> equal ranking of all the other options.

That's not true, or at least most people's deployment of Condorcet
_does_ allow equal ranking of all the other options. The most obvious
example is the Debian one. Why don't we just use the interpretation
of Condorcet specified nicely in the (newly-revised) Debian
Constitution ?

Ian.

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html;charset=ISO-8859-1">
<title></title>
</head>
<body text="#000000" bgcolor="#ffffff">
Indeed. The Debian project is SPI's main activity. That project has
dealt with this problem with lengthy and proper process, in a manner
that few would object to. Why not adopt their system?<br>
<br>
&nbsp;&nbsp;&nbsp; Thanks<br>
<br>
&nbsp;&nbsp;&nbsp; Bruce<br>
<br>
Ian Jackson wrote:<br>
<blockquote type="cite"
cite="mid16288(dot)1672(dot)848512(dot)571288(at)chiark(dot)greenend(dot)org(dot)uk">
<pre wrap="">Wichert Akkerman writes ("Re: Board of directors membership election"):
</pre>
<blockquote type="cite">
<pre wrap="">Previously Bdale Garbee wrote:
</pre>
<blockquote type="cite">
<pre wrap="">Treating such ballots as worthless seems overly pedantic to me.
</pre>
</blockquote>
<pre wrap="">Fact remains that you have to with condorcet since it doesn't allow
equal ranking of all the other options.
</pre>
</blockquote>
<pre wrap=""><!---->
That's not true, or at least most people's deployment of Condorcet
_does_ allow equal ranking of all the other options. The most obvious
example is the Debian one. Why don't we just use the interpretation
of Condorcet specified nicely in the (newly-revised) Debian
Constitution ?

Ian.

_______________________________________________
Spi-general mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Spi-general(at)lists(dot)spi-inc(dot)org">Spi-general(at)lists(dot)spi-inc(dot)org</a>
<a class="moz-txt-link-freetext" href="http://lists.spi-inc.org/cgi-bin/listinfo/spi-general">http://lists.spi-inc.org/cgi-bin/listinfo/spi-general</a>

</pre>
</blockquote>
</body>
</html>

Bruce Perens writes ("Re: Board of directors membership election"):
> <!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
> <html>
> <head>
> <meta http-equiv="Content-Type" content="text/html;charset=ISO-8859-1">
> <title></title>
> </head>
> <body text="#000000" bgcolor="#ffffff">
> Indeed. The Debian project is SPI's main activity. That project has
> dealt with this problem with lengthy and proper process, in a manner
> that few would object to. Why not adopt their system?<br>

Bruce, surely you should know better than to reply in HTML !

Ian.

Previously Ian Jackson wrote:
> Bruce, surely you should know better than to reply in HTML !

Especially since that at least triggered my spamfilters so I never saw
it.

Wichert.

--
Wichert Akkerman <wichert(at)wiggy(dot)net> It is simple to make things.
http://www.wiggy.net/ It is hard to make things simple.

OK, we are set to plain text + HTML now. I /like/ using italics and
*bold *once in a while. And here's a picture of my little boy.
<http://stanley.perens.com/>

Bruce

Wichert Akkerman wrote:

>Previously Ian Jackson wrote:
>
>
>>Bruce, surely you should know better than to reply in HTML !
>>
>>
>
>Especially since that at least triggered my spamfilters so I never saw
>it.
>
>Wichert.
>
>
>

Bruce Perens <bruce(at)perens(dot)com> wrote:

> OK, we are set to plain text + HTML now. I like using italics and bold
> once in a while. And here's a picture of my little boy.
> You'd be seeing Stanley's image here if you weren't living in a cave.
> Bruce

This is way off-topic, but you're quite wrong Bruce.

Your mixed message had three MIME parts:

- plain text version of the message
- HTML version of the same Message
- jpeg image

It's perfectly possible to view your message as intended with the inline
image without the HTML MIME part. You can simply send messages in plain
text and include images as MIME parts (either displayed inline or as
attachments). HTML is *not* needed to send an image.

Further, depending on your MUA, the word *not* above (and again here)
might be displayed as bold. All without using HTML. So I don't live
in a cave and I don't send HTML.

Hope this helps!

Peter

On Thu, Oct 30, 2003 at 10:47:10AM -0500, Peter S Galbraith wrote:
> - plain text version of the message
> - HTML version of the same Message
> - jpeg image
>
> It's perfectly possible to view your message as intended with the inline
> image without the HTML MIME part. You can simply send messages in plain
> text and include images as MIME parts (either displayed inline or as
> attachments). HTML is *not* needed to send an image.

What's more, if your image is represented by an <IMG> tag in HTML, many
people will not see it because of the security risks of downloading such
links, and of the requirements of being online at the time it is downloaded.

As an example, a spammer could insert a link:

<IMG SRC="http://spamserver.com/cgi-bin/validemail?good=foo(at)debian(dot)org">

The CGI could note the good message and pass back a GIF*.

Thus relying on HTML to transmit pictures is really not a good thing to do
at all.

* Of course it's a GIF. GIFs are evil and this is a spammer, right? :-)

> Further, depending on your MUA, the word *not* above (and again here)
> might be displayed as bold. All without using HTML. So I don't live
> in a cave and I don't send HTML.
>
> Hope this helps!
>
> Peter
>
> _______________________________________________
> Spi-general mailing list
> Spi-general(at)lists(dot)spi-inc(dot)org
> http://lists.spi-inc.org/cgi-bin/listinfo/spi-general

The board of directors of Software in the Public Interest has three
vacancies which need to be filled. At its last meeting the board
accepted resolution 2003-10-14.iwj.6 which regulates an election to
fill those positions and which temporarily reappointed Ian Jackson and
Martin Schulze to fill two vacancies for the next ninety days, or until
elected new members are installed, whichever comes earlier.

The election will be done using the same procedure as used for the
previous election. The voting procedure will be as follows:

* From October 23 to end of November 6 (UTC) contributing members
can declare their candidacy by submitting a position statement to
the SPI secretary at secretary(at)spi-inc(dot)org .

* From November 7 to end of November 21 (UTC) contributing members can
vote on the candidates.

Votes will be counted using the "Condorcet" election method system which
will be used to select the most preferred candidate. Conceptually, the
election will be broken into a a series of pairwise races between each
possible paring of the candidates. If one candidate beats each of the
others in pairwise races, that candidate wins. Otherwise, the
"Cloneproof/Schwartz Sequential Dropping" method is employed to choose
the most preferred candidate from those remaining. Some background
reading on preferential voting and Condorcet is available online at
http://www.electionmethods.org/ .

Wichert.

--
Wichert Akkerman <wichert(at)wiggy(dot)net> It is simple to make things.
http://www.wiggy.net/ It is hard to make things simple.