Charles Plessy <
plessy@debian.org> writes:
here are comments from my point of view in the hope of being useful,
please do not take them as objections; I know that you have looked at
the problem through many more angles than I did. By the way, sorry in advance if what I write today has been already addressed by others
earlier.
Oh, no, this is very helpful, and don't worry, you're not repeating
anything other folks have said.
Le Thu, Nov 11, 2021 at 08:26:00AM -0800, Russ Allbery a écrit :
Once a proposal has been sponsored and added to the ballot, we, as a
general social convention, stop sponsoring it unless it feels
particularly important to be listed as a sponsor.
This practice of sponsoring over the necessary number has always made me uncomfortable. The more we give importance to the question of who
sponsored, the more we put social pressure on the voters. This is
another reason why I think that limiting strictly to 5 would be better.
This said, anonymous voting (which I think is better to decide in a
separate GR) would solve that problem entirely.
I think this matches my opinion as well.
We do have a reasonably strong social convention to not pile on
sponsorships, which has been sufficient in recent GRs to keep the number
of sponsors at 15 or less. Which is still a lot more than 5, but which is still quite small compared to the total number of people who vote.
In other words, I think once a ballot option makes it onto the ballot,
the rules are attempting to capture the sense that it no longer belongs
just to its proposer, but now represents some unknown number of people
who want to vote for it.
My conclusion with the 2008 GR is that removing the original option made
it a better GR. The removal triggered the addition of new options, but
none of them were identical to the original one. Perhaps it would have happened also even if sponsors had a veto, but I believe that personal responsibility is going to be more efficient than collective
intelligence there.
This paragraph helped me realize that you see my proposal as a substantive change over the current rules for the original GR proposer. I hadn't
entirely realized that it was, and I see that I missed some subtlety in
current rules, so I went back and reviewed them. I believe the
implications of the current constitution (which I think was also in effect
in 2008) are:
1. The original proposer of the GR can only change the text themselves if
all the sponsors agree (A.1.5)
2. Someone else can propose a formal amendment to the text of the
resolution by getting the normal number of sponsors, in which case the
original proposer of the GR can accept it without needing the approval
of any other sponsors and it replaces the original text (A.1.1 and
A.1.2)
3. A formal amendment can be rejected by the original proposer of the GR,
at which point it becomes a new ballot option. At that point, only the
original proposer of the GR can suggest changes to the unaccepted
amendment, and all sponsors of the unaccepted amendment have to agree
for the change to be made (A.1.5). Otherwise, unaccepted amendments
can't be changed at all, but can be withdrawn (A.4).
The part that I was focused on fixing was the last part, which treats "unaccepted amendments" (the current term for ballot options) in a rather
weird way and gives the original GR proposer the sole power to modify them
even though, in practice, those amendments are often contrary to the
opinion of the original GR proposer. My intent was to make the rules in situation 1 apply to situation 3 and also get rid of this special and
somewhat odd grant of powers to the original proposer.
I think what you're pointing out is that 2 also changes in my proposal,
which I hadn't paid close attention to. Currently, if a change to the
original GR (but not any of the unaccepted amendments) has the support of
K+1 people (the proposer of the amendment plus the sponsors), the original proposer of the GR can choose to replace their text with the amended text without having to get the approval of the existing sponsors. Under my proposal, this concept of a sponsored amendment doesn't exist; if someone
wants to change a ballot option rather than add a new one, they have to convince the proposer of that ballot option and all of its sponsors.
I agree that this is a change; I'm eliminating the whole idea of an
accepted amendment and relying on only the first point. But I think it's
a relatively minor change that leads to a lot of process simplification
but not much difference in the outcome.
Under my proposed system, if there is K+1 support for a new form of a
ballot option, the original proposer agrees, but one of the original
sponsors disagrees, the (slightly more complex in this edge case) process
is:
1. The new ballot option is proposed and sponsored as normal. The
original GR proposer can sponsor that new ballot option. It is added
to the ballot.
2. The original GR proposer withdraws as proposer of the original ballot
option. This triggers its removal from the ballot in 24 hours unless
someone else steps forward as proposer and it still has enough
sponsors. Presumably the original sponsor who disagrees will step
forward to become the proposer, but they will need to retain the
support of the other sponsors (the other sponsors will need to not
withdraw and, if there were only five, another sponsor will have to
step forward), or alternatively replace them with new sponsors.
My argument is that this is mostly equivalent. The new option is put on
the ballot, the original proposer shifts their support to the new option,
and the old option is removed unless K+1 people continue to support it.
That said, the place where this is not equivalent is that it is somewhat
easier to keep the old form of the option on the ballot in my proposed
system because that support can be passive rather than active (only one or
two people have to proactively do something to keep it on the ballot,
instead of six people). I think the question is whether that makes enough
of a difference to warrant adding additional complexity here.
Fraknly speaking, I worry that some day somebody will sponsor as many
options as possible and object to changes for the sake of adding
toxicity to GR. The more members we have, the less unlikely it becomes.
I think this is possible but I'm not sure that it matters. This is where
Sam's observation that the option change process is effectively an
optimization for proposing a new one and withdrawing the old one comes
into play.
If it's only one person doing this, the result will be some tedium as
people withdraw from the previous version of the option and add their
support to the new version of the option, but we'll still get the same
ballot in the end because the old option can't be kept on the ballot by a single person.
The major drawback is that it requires K+1 people be actively involved in
the process to make all of that happen, but in the case where someone is
doing this maliciously, I don't think it will be a problem to find people
to counter that.
And, if we end up with a few spurious options on the ballot, then in
theory this is what having a clone-proof voting system is for.
But even assuming good will, I think that a process that makes it easier
to remove or merge options would serve us better than a process that
makes it easier to fork options. A lot of GR voters do not follow the debates on debian-vote, and with too many options, possibly all written
in different styles, there is an increased risk of voting for the
contrary of what we want.
My feeling on this is that some of the remedy is social. If one is
following debian-vote and participating in the process, one can help make
the ballot clearer by sponsoring better versions of options and
withdrawing support from less-clear versions of options, and that will
achieve removal and merging of options as long as there aren't K+1 people
who disagree.
That said, I think it is true (and intentional) in both our current system
and in my system that if K+1 people want something on the ballot, it will
be on the ballot, and this is a very low bar. In other words, both the
old and the new system are erring on the side of having more options on
the ballot and relying on the fact that our voting system is clone-proof.
Some of the objections that I've seen after recent GRs to complex ballots
are actually objections to relying on the clone-proof nature of the voting system, or arguments that because of human psychology it's not really clone-proof because people get confused. I admit to being somewhat
dubious of those arguments (they're often based on the supposed unreasonableness of voting patterns that I find entirely reasonable and rational), but, more to the point for this proposal, I think I'm going to
plead "out of scope" for the changes that I'm trying to make. I don't
think this alteration of the process will make the problem of having too
many ballot options any worse (and hopefully the above is a convincing
argument for why I feel that way), and I'd rather leave trying to make it better as a subject for another GR.
--
Russ Allbery (
rra@debian.org) <
https://www.eyrie.org/~eagle/>
--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)