inamerrata

In late November, I did up a proposal to provide a way of making some of the interesting posts hidden away on the developer-only debian-private mailing list more public — mostly on the basis that secret discussions aren’t good for Debian, and that there’s some really fascinating discussions on topics which continue to come up, that I’d like to be able to refer to directly.

In the end, that didn’t pass, although we instead ended up with a compromise position whereby we won’t release any of the old posts to that list, only the new ones; and there won’t be any practical result from the GR for another three years.

Since this wasn’t a leadership election, who voted for what is now public, so we can do a few interesting analyses. At the most basic, there were five major categories of voters:

• 89 people who wanted as much declassification as possible (V: 123)
• 54 people who wanted as little declassification as possible (V: 321)
• 50 people who would accept future declassifaction, but not past (V: 312)
• 38 people who preferred future declassification (V: 213)
• 37 people who didn’t want declassification, but had no other preference (V: 221)

The 30 voters who didn’t fit into one of the other categories pushed past declassification past further discussion (19:2), and closed the gap between past and future disclosure (19:8). As it turns out, IRV (the preferential voting system used in Australia) would still have produced the same result, so I think we’re still yet to come up with an example where Condorcet acts differently.

An interesting question is whether this result’s a good compromise or not; the two most common sets of votes were exactly opposite, and the next most common was in favour of the compromise — which is exactly the right case for our voting system to produce a compromise result. OTOH, that means 202 of the 298 voters (a 2:1 supermajority) didn’t get their first preference and end up dissatisfied. I wonder if, in a voluntary organisation, it makes sense to leave most people dissatisfied like that, or to take a polarised result and leave the people who end up strongly dissatisfied to move on to other things. I also wonder how that’ll look in a few years, or even a few months: at that point a tentative compromise on this issue might seem entirely sensible and mature, and any more radical action seem absurdly headstrong.

Speaking of compromise, one thing that’s interesting is what effect the various compromises made while the proposal was being discussed actually had. There were three significant ones that got incorporated into the main proposal:

• Author veto suggested by Manoj Srivastava
• Simpler delaying process suggested by Don Armstrong
• No exceptions for security problems suggested by Bernhard Link

In addition, an alternative proposal to only deal with future posts was made in response to concerns expressed by John Lightsey and MJ Ray. Of those, Don Armstrong voted in favour of declassification (123), and John Lightsey voted in favour of future declassification only (312), while both Manoj and Bernhard voted against declassifciation (321), and MJ appears not to have voted at all.

Of the sponsors of the main proposal, all voted for declassifcation (123); of the four sponsors of the alternate proposal who didn’t also sponsor the main proposal, two then voted against declassifaction (Gregory Norris, David Garza; 321), leaving only Neil McGovern (213) and John Lightsey voting for the proposal as well as sponsoring it.

I’m not sure what all that actually means, but it does leave me wondering how much point there actually is to trying to address the concerns raised by people on the lists. Which is a bit of a let down, since that’s what I find most entertaining about Debian.

As usual, the votes came in blocks: the first hundred in the first day and a half after the initial call for votes, the next fifty over the following week, another twenty-five after the second call for votes, and another fifty in the week following up until the final call for votes, which generated an additional seventy-five or so votes. So day one and two, and thirteen and fourteen accounted for 29% of the time, but 59% of the vote. It’d be interesting to see how well (or badly) a vote worked that only lasted four days, perhaps with some sort of “postal vote” arrangement so people who are going to be away can provisionally vote while discussion is still ongoing.

Three hundred voters was actually a pretty good turn out — coming to 30.8%. In spite of being mostly over the Christmas/New Year period, that beats the turnout for most of the other elections we’ve had. The consitution had a 24% turnout, the logo vote had 28.1%, the logo swap had 21.2%, the logo license had 21.5%, the condorcet vote had 19.9%, the allow changes to the social contract vote had 26%, and the editorial changes vote had 23.5%. That leaves the only votes to have higher turnout being the non-free vote at 53.1%, the sarge release followup vote at 43.6%, and the DPL votes which ranged from 50.6% (2002) to 62.2% (2000).