>>>>> "Charles" == Charles Plessy <plessy@debian.org> writes:
Charles> One last question: in some complex GRs there were
Charles> discussions about problems caused by mixing 1:1 and 3:1
Charles> majority options, which frankly speaking I could not
Charles> undertand because I never studied our Condorcet method in
Charles> details. Do you think that such mixes can be problemating
Charles>
I can explain the issues I see.
Whether they are problematic depends on how you think things ought to
work.
1) Debian prefers an answer to FD.
So, consider the following options:
1) Change the DFSG.
2) The DFSG is great
3) FD
Option 1 defeats option 2
Option 1 defeats option 3 by say 2.0:1
So, option 1 cannot win because option 1 needs to defeat option 3 by
3:1 to win.
There are two reasonable things we could have said the rules cause to
happen in that case. First, we could have said that if an option would
have won but for super majority, then inherently FD wins.
Or we could (and did) say that option gets dropped.
So in the situation above, "The DFSG is great" wins even though more
people would have preferred to replace the DFSG than to say it was
great.
I've intentionally phrased things to maximize how bad it sounds.
If we allowed FD to win in this situation, it would be open to strategic
abuse: you could potentially drag out the discussion by introducing a
supermajority option that you thought would win, but not win enough.
But it also seems like the current system is open to abuse because of
the situation I described above.
Whether that's true depends on how you think about FD.
Consider another restatement of the results above.
Most people are either okay revising the DFSG or saying the DFSG is
great.
Both options are fine with the project.
More people would prefer to revise the DFSG, but not enough people to
actually permit the change.
We prefer to be done with discussions rather than have them drag out,
and we've encoded that to a certain extent in our constitution.
So, we decided to say the DFSG was great because we could do that today
rather than let things drag out.
2) There's another issue involving supermajorities.
In a ballot like the one we're about to face, I might have an incentive
to rank an opposing constitutional amendment below FD even though I'd
rather see that amendment succeed than have more discussion. After all
it might fail its super majority if enough people do that.
Again, whether this is abuse depends a lot on how you think about voter
intent.
The only way to avoid all of these issues is to avoid super majorities
entirely.
We're effectively saying that there are cases where the voters prefer
some option, and we're not going to let the voters choose that option
because they don't prefer it enough.
I think that is going to open up some strategic abuse somewhere.
The question is what sort of abuse do you permit.
Russ's proposal does not make any changes in these areas.
I actually think we have a good set of trade offs already and so I'm
happy with that.
But others might like a different set of trade offs.
Attachment:
signature.asc
Description: PGP signature