You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@airflow.apache.org by Daniel Standish via dev <de...@airflow.apache.org> on 2025/10/22 21:25:29 UTC
[DISCUSS] multiple choice voting procedure
The DAG terminology vote I think has surfaced a problem with our multiple
choice voting procedure.
If you allow people to vote for multiple options, they seem to tend to use
it in a manner to signify their ranked preference. However, this could
easily result in an option that doesn't have majority preference getting
the win.
E.g. suppose 4 people vote for option A, and 5 people vote for option B +1
but also +0.5 for A. Then option A will win even though people prefer
option B 5 to 4.
This is a bad outcome.
It gets even stranger if you allow negative votes. Then you end
essentially invalidating other peoples votes, unless *everyone* minuses all
of the options they don't favor. And even if everyone does that, then it's
hard to see how that gets to the outcome favored by most.
With ranked choice voting, everyone votes for their most favored choice,
but they can also rank all the options. If their most favored option does
not win, then their vote goes to their second favored option, and so on.
This is a better way to do this.
I propose that when doing multiple choice votes, we do ranked choice,
instead of allowing people to just vote for multiple options with plus or
minus votes.
Re: [DISCUSS] multiple choice voting procedure
Posted by Tzu-ping Chung via dev <de...@airflow.apache.org>.
You don’t need negative votes to express concerns in a multiple choice election. This is how real-world elections work as well.
Instead, you can just express the concerns without casting votes, and if those concerns are agreed on, the problematic choice can be removed, or the entire voting is called off.
This is what -1 actually means in Apache Voting Process anyway, except the system is designed for an approval context, and the only choice being removed always results in the process being called off.
Of course, we can still “allow” -1 but keep the meaning as “please remove this choice.” But IMO that would be even more confusing (since negative votes would work differently from positive votes).
TP
> On 23 Oct 2025, at 11:49, Kiruban Kamaraj <ki...@gmail.com> wrote:
>
> IMHO, if we don't let people vote -1, how are devs supposed to raise
> legitimate concerns about an option? If someone votes -1, they should
> explain why - and hopefully devs are only doing this when they have real
> concerns, not just to push their favorite choice. If you have a solid
> reason to oppose something, just vote -1. If you don't care about another
> option either way, then don't vote on it.
>
> I agree with what Jerek said.
>
> *it's not "who wins" but "which option wins". I don't absolutely care who*
> *"wins" here, but which option has the most support.*
>
> On Thu, Oct 23, 2025 at 2:55 AM Daniel Standish via dev <
> dev@airflow.apache.org> wrote:
>
>> The DAG terminology vote I think has surfaced a problem with our multiple
>> choice voting procedure.
>>
>> If you allow people to vote for multiple options, they seem to tend to use
>> it in a manner to signify their ranked preference. However, this could
>> easily result in an option that doesn't have majority preference getting
>> the win.
>>
>> E.g. suppose 4 people vote for option A, and 5 people vote for option B +1
>> but also +0.5 for A. Then option A will win even though people prefer
>> option B 5 to 4.
>>
>> This is a bad outcome.
>>
>> It gets even stranger if you allow negative votes. Then you end
>> essentially invalidating other peoples votes, unless *everyone* minuses all
>> of the options they don't favor. And even if everyone does that, then it's
>> hard to see how that gets to the outcome favored by most.
>>
>> With ranked choice voting, everyone votes for their most favored choice,
>> but they can also rank all the options. If their most favored option does
>> not win, then their vote goes to their second favored option, and so on.
>>
>> This is a better way to do this.
>>
>> I propose that when doing multiple choice votes, we do ranked choice,
>> instead of allowing people to just vote for multiple options with plus or
>> minus votes.
>>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@airflow.apache.org
For additional commands, e-mail: dev-help@airflow.apache.org
Re: [DISCUSS] multiple choice voting procedure
Posted by Jarek Potiuk <ja...@potiuk.com>.
I'd say we can strive for a single method of multi-choice. It seems - from
the earlier discussion - that there are many properties of "simple" voting
that make people worried. That's a valid worry - and if we already have a
way (I assume we will use some very simple tooling) to do IRV - IRV voting
is "way simpler" for the voting people. You do not have to think what +1,
-1, 0 means - you do not have to worry about how your -1 will be perceived,
the only thing you have to do is order the choices according to your
preferences. Calculation is a bit more complex, but well, we have engineers
and there are tools available - so I guess we can manage it.
I think also - if we agree to it upfront - it helps with just "following
the book". The case we have with Dag naming - was a good example - in the
middle of voting concerns were raised that things are unclear and that
there are doubts about the way of voting. Could those concerns be raised
before - surely, but they weren't and we had to start discussing mid-voting
That has actually a socially-intimidating effect - it suggests (even if not
intentionally) that the voting process was not well thought out and is not
really good, and that we should have done better (even if we've "always
done it this way").
I think if we leave the option of "simple voting" on the table, this might
repeat the story - we might start voting and someone in the middle of it
will start questioning the process. I strongly prefer that we set up the
process - and even have an example voting done "well" - i.e. where we
generally agree on the process and follow it and do not leave the option
that one of us starts questioning it in the middle. No matter the intention
- this is pretty disruptive. So I would strongly prefer to have just one
way of voting on multiple options - and seems that we might even get a
consensus on that (IRV) and simply call for LAZY CONSENSUS on using it in
the future for all such votings.
Also - I think it would be a good exercise to complete current Dag voting
and re-assume it using the IRV method - while we are still focused on that
subject - even if just out of curiosity what will be the difference in
results. And that gives an opportunity to see if our new process works, get
some tooling to do it and see if we all understand how it works. Also it
might give those who see deficiencies in the current process to run them
themselves and show how it should look like - with good communication of
goals, options and process. It's always easier to say "I would like to
change the process", but the real deal (and way better) is to "run it
yourself" - to show how it should be done. That also shows real commitment
to changing the process.
J.
On Fri, Oct 24, 2025 at 7:58 AM Amogh Desai <am...@apache.org> wrote:
> Sounds like a good plan, Daniel.
>
>
> Thanks & Regards,
> Amogh Desai
>
>
> On Thu, Oct 23, 2025 at 9:53 PM Daniel Standish via dev <
> dev@airflow.apache.org> wrote:
>
> > Yes IRV is what I was referring to when I write ranked choice vote. If
> > that's the approach that Apache favors / supports, then it sounds good to
> > me.
> >
> > As I mentioned today in the other thread, I think it is generally better
> if
> > we can avoid multiple choice votes. But when it makes sense, then IRV
> > seems right to me.
> >
> > I think it also could be reasonable to start out with an initial vote of
> > just simple one option selection, i.e. non-transferrable vote and
> majority
> > wins. And then if it's close, anybody can request we do IRV. The reason
> > for this is it might not be close enough to matter and it's simpler to
> not
> > do IRV.
> >
> > But I think we should not allow voting for multiple options because it's
> > confusing, hard to reason about, and as mentioned in the original message
> > of this thread, you may be unwittingly voting against your interest.
> >
> >
> > So my proposal I think would be the following two provisions:
> >
> > 1. For multiple choice votes we have two options:
> > a. simple majority vote, each voter gets one non-transferrable vote, and
> > cannot select multiple options
> > b. IRV
> >
> > 2. The person calling the vote can choose which voting approach to use.
> If
> > simple vote is chosen, any community member eligible to vote can request
> > changing to IRV for any reason.
> >
> >
> >
> >
> > On Thu, Oct 23, 2025 at 1:02 AM Jarek Potiuk <ja...@potiuk.com> wrote:
> >
> > > > Oh absolutely - sorry I missed it :)
> > >
> > > Actually I started to do research and write my response before you
> > > responded - so the issue is that I did not have a chance to read your
> > > message :) ... Awfully sorry for that Amogh... Mental note to myself -
> > good
> > > thing to read the new answers before you press send on yours ;)
> > >
> > > On Thu, Oct 23, 2025 at 9:54 AM Jarek Potiuk <ja...@potiuk.com> wrote:
> > >
> > > >
> > > >
> > > > On Thu, Oct 23, 2025 at 9:50 AM Amogh Desai <am...@apache.org>
> > > wrote:
> > > >
> > > >> Yeah, I was proposing Instant Runoff earlier when I said this:
> > > >>
> > > >> > I would second Daniel to have a rank based voting for ballots
> which
> > > can
> > > >> have multiple choice:
> > > https://en.wikipedia.org/wiki/Instant-runoff_voting
> > > >> .
> > > >>
> > > >
> > > > Oh absolutely - sorry I missed it :) - there were too many emails in
> > the
> > > > thread to read this morning and I simply missed yours, sorry Amogh.
> > And I
> > > > picked my proposal even without realising you had the same proposal -
> > > > which might mean it is a really good idea :)
> > > >
> > > > And one more comment:
> > > >
> > > > > I don’t necessarily think this is a bad outcome. Those 5 people
> also
> > > > believe A is a good idea, just not as good as B.
> > > >
> > > > Yes. I agree with Wei Lee here. It's not a bad outcome. It's SOME
> > > outcome.
> > > > It has some properties (and also some social aspects to it - like
> > people
> > > > are less likely to put -1 even if they do not agree with something
> and
> > > few
> > > > other things. Again - it disregards individual voting power of a
> single
> > > > person (if the voting power is something we seek as desired
> property),
> > it
> > > > focuses more on the single option support, not whether one person has
> > > > always the same "strength" of their vote. It assumes that we have
> > > > collaborating people who will not seek to 'play" the system but
> simply
> > > > state their preferences. It's not tamper-proof and if we have bad
> > > players,
> > > > they can definitely abuse it.
> > > >
> > > > So yes - if we are at the stage where we are concerned about bad
> actors
> > > > trying to play the system to their advantage, then yes, we should
> > choose
> > > a
> > > > system that is tamper proof. To be honest that never occured to me in
> > the
> > > > past that we are at this stage, but it's a good idea anyway.
> > > >
> > >
> >
>
Re: [DISCUSS] multiple choice voting procedure
Posted by Amogh Desai <am...@apache.org>.
Sounds like a good plan, Daniel.
Thanks & Regards,
Amogh Desai
On Thu, Oct 23, 2025 at 9:53 PM Daniel Standish via dev <
dev@airflow.apache.org> wrote:
> Yes IRV is what I was referring to when I write ranked choice vote. If
> that's the approach that Apache favors / supports, then it sounds good to
> me.
>
> As I mentioned today in the other thread, I think it is generally better if
> we can avoid multiple choice votes. But when it makes sense, then IRV
> seems right to me.
>
> I think it also could be reasonable to start out with an initial vote of
> just simple one option selection, i.e. non-transferrable vote and majority
> wins. And then if it's close, anybody can request we do IRV. The reason
> for this is it might not be close enough to matter and it's simpler to not
> do IRV.
>
> But I think we should not allow voting for multiple options because it's
> confusing, hard to reason about, and as mentioned in the original message
> of this thread, you may be unwittingly voting against your interest.
>
>
> So my proposal I think would be the following two provisions:
>
> 1. For multiple choice votes we have two options:
> a. simple majority vote, each voter gets one non-transferrable vote, and
> cannot select multiple options
> b. IRV
>
> 2. The person calling the vote can choose which voting approach to use. If
> simple vote is chosen, any community member eligible to vote can request
> changing to IRV for any reason.
>
>
>
>
> On Thu, Oct 23, 2025 at 1:02 AM Jarek Potiuk <ja...@potiuk.com> wrote:
>
> > > Oh absolutely - sorry I missed it :)
> >
> > Actually I started to do research and write my response before you
> > responded - so the issue is that I did not have a chance to read your
> > message :) ... Awfully sorry for that Amogh... Mental note to myself -
> good
> > thing to read the new answers before you press send on yours ;)
> >
> > On Thu, Oct 23, 2025 at 9:54 AM Jarek Potiuk <ja...@potiuk.com> wrote:
> >
> > >
> > >
> > > On Thu, Oct 23, 2025 at 9:50 AM Amogh Desai <am...@apache.org>
> > wrote:
> > >
> > >> Yeah, I was proposing Instant Runoff earlier when I said this:
> > >>
> > >> > I would second Daniel to have a rank based voting for ballots which
> > can
> > >> have multiple choice:
> > https://en.wikipedia.org/wiki/Instant-runoff_voting
> > >> .
> > >>
> > >
> > > Oh absolutely - sorry I missed it :) - there were too many emails in
> the
> > > thread to read this morning and I simply missed yours, sorry Amogh.
> And I
> > > picked my proposal even without realising you had the same proposal -
> > > which might mean it is a really good idea :)
> > >
> > > And one more comment:
> > >
> > > > I don’t necessarily think this is a bad outcome. Those 5 people also
> > > believe A is a good idea, just not as good as B.
> > >
> > > Yes. I agree with Wei Lee here. It's not a bad outcome. It's SOME
> > outcome.
> > > It has some properties (and also some social aspects to it - like
> people
> > > are less likely to put -1 even if they do not agree with something and
> > few
> > > other things. Again - it disregards individual voting power of a single
> > > person (if the voting power is something we seek as desired property),
> it
> > > focuses more on the single option support, not whether one person has
> > > always the same "strength" of their vote. It assumes that we have
> > > collaborating people who will not seek to 'play" the system but simply
> > > state their preferences. It's not tamper-proof and if we have bad
> > players,
> > > they can definitely abuse it.
> > >
> > > So yes - if we are at the stage where we are concerned about bad actors
> > > trying to play the system to their advantage, then yes, we should
> choose
> > a
> > > system that is tamper proof. To be honest that never occured to me in
> the
> > > past that we are at this stage, but it's a good idea anyway.
> > >
> >
>
Re: [DISCUSS] multiple choice voting procedure
Posted by Daniel Standish via dev <de...@airflow.apache.org>.
Yes IRV is what I was referring to when I write ranked choice vote. If
that's the approach that Apache favors / supports, then it sounds good to
me.
As I mentioned today in the other thread, I think it is generally better if
we can avoid multiple choice votes. But when it makes sense, then IRV
seems right to me.
I think it also could be reasonable to start out with an initial vote of
just simple one option selection, i.e. non-transferrable vote and majority
wins. And then if it's close, anybody can request we do IRV. The reason
for this is it might not be close enough to matter and it's simpler to not
do IRV.
But I think we should not allow voting for multiple options because it's
confusing, hard to reason about, and as mentioned in the original message
of this thread, you may be unwittingly voting against your interest.
So my proposal I think would be the following two provisions:
1. For multiple choice votes we have two options:
a. simple majority vote, each voter gets one non-transferrable vote, and
cannot select multiple options
b. IRV
2. The person calling the vote can choose which voting approach to use. If
simple vote is chosen, any community member eligible to vote can request
changing to IRV for any reason.
On Thu, Oct 23, 2025 at 1:02 AM Jarek Potiuk <ja...@potiuk.com> wrote:
> > Oh absolutely - sorry I missed it :)
>
> Actually I started to do research and write my response before you
> responded - so the issue is that I did not have a chance to read your
> message :) ... Awfully sorry for that Amogh... Mental note to myself - good
> thing to read the new answers before you press send on yours ;)
>
> On Thu, Oct 23, 2025 at 9:54 AM Jarek Potiuk <ja...@potiuk.com> wrote:
>
> >
> >
> > On Thu, Oct 23, 2025 at 9:50 AM Amogh Desai <am...@apache.org>
> wrote:
> >
> >> Yeah, I was proposing Instant Runoff earlier when I said this:
> >>
> >> > I would second Daniel to have a rank based voting for ballots which
> can
> >> have multiple choice:
> https://en.wikipedia.org/wiki/Instant-runoff_voting
> >> .
> >>
> >
> > Oh absolutely - sorry I missed it :) - there were too many emails in the
> > thread to read this morning and I simply missed yours, sorry Amogh. And I
> > picked my proposal even without realising you had the same proposal -
> > which might mean it is a really good idea :)
> >
> > And one more comment:
> >
> > > I don’t necessarily think this is a bad outcome. Those 5 people also
> > believe A is a good idea, just not as good as B.
> >
> > Yes. I agree with Wei Lee here. It's not a bad outcome. It's SOME
> outcome.
> > It has some properties (and also some social aspects to it - like people
> > are less likely to put -1 even if they do not agree with something and
> few
> > other things. Again - it disregards individual voting power of a single
> > person (if the voting power is something we seek as desired property), it
> > focuses more on the single option support, not whether one person has
> > always the same "strength" of their vote. It assumes that we have
> > collaborating people who will not seek to 'play" the system but simply
> > state their preferences. It's not tamper-proof and if we have bad
> players,
> > they can definitely abuse it.
> >
> > So yes - if we are at the stage where we are concerned about bad actors
> > trying to play the system to their advantage, then yes, we should choose
> a
> > system that is tamper proof. To be honest that never occured to me in the
> > past that we are at this stage, but it's a good idea anyway.
> >
>
Re: [DISCUSS] multiple choice voting procedure
Posted by Jarek Potiuk <ja...@potiuk.com>.
> Oh absolutely - sorry I missed it :)
Actually I started to do research and write my response before you
responded - so the issue is that I did not have a chance to read your
message :) ... Awfully sorry for that Amogh... Mental note to myself - good
thing to read the new answers before you press send on yours ;)
On Thu, Oct 23, 2025 at 9:54 AM Jarek Potiuk <ja...@potiuk.com> wrote:
>
>
> On Thu, Oct 23, 2025 at 9:50 AM Amogh Desai <am...@apache.org> wrote:
>
>> Yeah, I was proposing Instant Runoff earlier when I said this:
>>
>> > I would second Daniel to have a rank based voting for ballots which can
>> have multiple choice: https://en.wikipedia.org/wiki/Instant-runoff_voting
>> .
>>
>
> Oh absolutely - sorry I missed it :) - there were too many emails in the
> thread to read this morning and I simply missed yours, sorry Amogh. And I
> picked my proposal even without realising you had the same proposal -
> which might mean it is a really good idea :)
>
> And one more comment:
>
> > I don’t necessarily think this is a bad outcome. Those 5 people also
> believe A is a good idea, just not as good as B.
>
> Yes. I agree with Wei Lee here. It's not a bad outcome. It's SOME outcome.
> It has some properties (and also some social aspects to it - like people
> are less likely to put -1 even if they do not agree with something and few
> other things. Again - it disregards individual voting power of a single
> person (if the voting power is something we seek as desired property), it
> focuses more on the single option support, not whether one person has
> always the same "strength" of their vote. It assumes that we have
> collaborating people who will not seek to 'play" the system but simply
> state their preferences. It's not tamper-proof and if we have bad players,
> they can definitely abuse it.
>
> So yes - if we are at the stage where we are concerned about bad actors
> trying to play the system to their advantage, then yes, we should choose a
> system that is tamper proof. To be honest that never occured to me in the
> past that we are at this stage, but it's a good idea anyway.
>
Re: [DISCUSS] multiple choice voting procedure
Posted by Jarek Potiuk <ja...@potiuk.com>.
On Thu, Oct 23, 2025 at 9:50 AM Amogh Desai <am...@apache.org> wrote:
> Yeah, I was proposing Instant Runoff earlier when I said this:
>
> > I would second Daniel to have a rank based voting for ballots which can
> have multiple choice: https://en.wikipedia.org/wiki/Instant-runoff_voting.
>
Oh absolutely - sorry I missed it :) - there were too many emails in the
thread to read this morning and I simply missed yours, sorry Amogh. And I
picked my proposal even without realising you had the same proposal -
which might mean it is a really good idea :)
And one more comment:
> I don’t necessarily think this is a bad outcome. Those 5 people also
believe A is a good idea, just not as good as B.
Yes. I agree with Wei Lee here. It's not a bad outcome. It's SOME outcome.
It has some properties (and also some social aspects to it - like people
are less likely to put -1 even if they do not agree with something and few
other things. Again - it disregards individual voting power of a single
person (if the voting power is something we seek as desired property), it
focuses more on the single option support, not whether one person has
always the same "strength" of their vote. It assumes that we have
collaborating people who will not seek to 'play" the system but simply
state their preferences. It's not tamper-proof and if we have bad players,
they can definitely abuse it.
So yes - if we are at the stage where we are concerned about bad actors
trying to play the system to their advantage, then yes, we should choose a
system that is tamper proof. To be honest that never occured to me in the
past that we are at this stage, but it's a good idea anyway.
Re: [DISCUSS] multiple choice voting procedure
Posted by Amogh Desai <am...@apache.org>.
Yeah, I was proposing Instant Runoff earlier when I said this:
> I would second Daniel to have a rank based voting for ballots which can
have multiple choice: https://en.wikipedia.org/wiki/Instant-runoff_voting.
It is a more formal, well known and accepted form of voting for
multiple choices.
Thanks & Regards,
Amogh Desai
On Thu, Oct 23, 2025 at 1:17 PM Jarek Potiuk <ja...@potiuk.com> wrote:
> BTW. I think Instant Runoff Voting is kinda similar to what Daniel proposed
> but it's well researched and formalized way of doing it and it has known
> properties that you can read about in the wiki page.
>
> On Thu, Oct 23, 2025 at 9:40 AM Jarek Potiuk <ja...@potiuk.com> wrote:
>
> > If we are not happy with the current system that disregards voting power
> > and simply states for preferences on options - then yes we can change
> it .
> > Actually this is an important part of governance - to change the rules if
> > we are unhappy with them.
> >
> > And Actually we don't have to design our own system. ASF has us
> > covered.Voting types are a very well researched subject and there are
> > multiple voting systems we can choose from see "Electoral systems": [1] -
> > each of them with different properties. There were plenty of similar
> > discussions in the past in either ASF or ASF project that went the same
> way
> > - not happy with simple system, implementing more complex one that had
> > different "fairness" properties - and well we can pick one for
> > multi-voting options if we want, I would not encourage us figuring out
> our
> > own "approach" but simply pick from existing ones.
> >
> > ASF has a project called Apache Steve [2] that implements multiple voting
> > methods that ASF recommends. You can read about those methods at STeVe
> > page. We use it for board elections and we used it (and Single
> > Transferable Vote [3] method) this year to select the new logo
> candidates.
> > This is a "multi-winner" system - i.e. one that allows several winners.
> > There is a single-winner variant of it - Instant Runoff Voting [4] that
> can
> > be used instead - where you put your options in the sequence of
> preferences.
> > Generally - the Single Transferable Vote is widely considered in ASF as
> > the most "fair" and we use it for almost everything where multiple
> options
> > are possible.
> >
> > Note that we do not have to use it of course (though we could ask infra
> to
> > set it up for us) - this system is designed in the way that people have
> > one-time ballots and that the voting is anonymous and can't be really
> > traced to a person - and I think we are nowhere near that expectation.
> But
> > we can use one of the methods it implements. In simple voting like ours,
> it
> > would be rather easy to implement it manually in an shared spreadsheet
> (or
> > we can find a ready-solution that is less complex than STeVe)
> >
> > *My proposal is not to invent our own - but to use Instant Runoff Voting
> > (as the variant of the Apache-recommended STV).*
> >
> > More details on the Instant Runoff voting:
> >
> > People instead of voting +1, 0, -1 simply order the votes in preference.
> > Then one or more eliminations are used to simulate multiple runoff
> > elections. In each round, the candidate with the fewest first-preference
> > votes (among the remaining candidates) is eliminated. This continues
> until
> > only one candidate is left.
> >
> > [1] Electoral system: https://en.wikipedia.org/wiki/Electoral_system
> > [2] ASF Steve project: https://steve.apache.org/
> > [3] Single Transferable Vote
> > https://en.wikipedia.org/wiki/Single_transferable_vote
> > [4] Instant Runoff-voting
> > https://en.wikipedia.org/wiki/Instant-runoff_voting
> >
> > J.
> >
> >
> > On Thu, Oct 23, 2025 at 8:49 AM Amogh Desai <am...@apache.org>
> wrote:
> >
> >> I think I agree mostly with what Daniel says.
> >>
> >> We should arrive at a preference based voting to eliminate any sort of
> >> voting power bias.
> >>
> >> I would second Daniel to have a rank based voting for ballots which can
> >> have multiple choice:
> https://en.wikipedia.org/wiki/Instant-runoff_voting
> >> .
> >>
> >> The ad-hoc +/- voting has more problems than we can imagine, and those
> >> can be solved by ranked ballots -- giving everyone equal power while
> also
> >> being easy to explain and implement.
> >>
> >> Thanks & Regards,
> >> Amogh Desai
> >>
> >>
> >> On Thu, Oct 23, 2025 at 9:20 AM Kiruban Kamaraj <
> kirubankamaraj@gmail.com
> >> >
> >> wrote:
> >>
> >> > IMHO, if we don't let people vote -1, how are devs supposed to raise
> >> > legitimate concerns about an option? If someone votes -1, they should
> >> > explain why - and hopefully devs are only doing this when they have
> real
> >> > concerns, not just to push their favorite choice. If you have a solid
> >> > reason to oppose something, just vote -1. If you don't care about
> >> another
> >> > option either way, then don't vote on it.
> >> >
> >> > I agree with what Jerek said.
> >> >
> >> > *it's not "who wins" but "which option wins". I don't absolutely care
> >> who*
> >> > *"wins" here, but which option has the most support.*
> >> >
> >> > On Thu, Oct 23, 2025 at 2:55 AM Daniel Standish via dev <
> >> > dev@airflow.apache.org> wrote:
> >> >
> >> > > The DAG terminology vote I think has surfaced a problem with our
> >> multiple
> >> > > choice voting procedure.
> >> > >
> >> > > If you allow people to vote for multiple options, they seem to tend
> to
> >> > use
> >> > > it in a manner to signify their ranked preference. However, this
> >> could
> >> > > easily result in an option that doesn't have majority preference
> >> getting
> >> > > the win.
> >> > >
> >> > > E.g. suppose 4 people vote for option A, and 5 people vote for
> option
> >> B
> >> > +1
> >> > > but also +0.5 for A. Then option A will win even though people
> prefer
> >> > > option B 5 to 4.
> >> > >
> >> > > This is a bad outcome.
> >> > >
> >> > > It gets even stranger if you allow negative votes. Then you end
> >> > > essentially invalidating other peoples votes, unless *everyone*
> >> minuses
> >> > all
> >> > > of the options they don't favor. And even if everyone does that,
> then
> >> > it's
> >> > > hard to see how that gets to the outcome favored by most.
> >> > >
> >> > > With ranked choice voting, everyone votes for their most favored
> >> choice,
> >> > > but they can also rank all the options. If their most favored
> option
> >> > does
> >> > > not win, then their vote goes to their second favored option, and so
> >> on.
> >> > >
> >> > > This is a better way to do this.
> >> > >
> >> > > I propose that when doing multiple choice votes, we do ranked
> choice,
> >> > > instead of allowing people to just vote for multiple options with
> >> plus or
> >> > > minus votes.
> >> > >
> >> >
> >>
> >
>
Re: [DISCUSS] multiple choice voting procedure
Posted by Jarek Potiuk <ja...@potiuk.com>.
BTW. I think Instant Runoff Voting is kinda similar to what Daniel proposed
but it's well researched and formalized way of doing it and it has known
properties that you can read about in the wiki page.
On Thu, Oct 23, 2025 at 9:40 AM Jarek Potiuk <ja...@potiuk.com> wrote:
> If we are not happy with the current system that disregards voting power
> and simply states for preferences on options - then yes we can change it .
> Actually this is an important part of governance - to change the rules if
> we are unhappy with them.
>
> And Actually we don't have to design our own system. ASF has us
> covered.Voting types are a very well researched subject and there are
> multiple voting systems we can choose from see "Electoral systems": [1] -
> each of them with different properties. There were plenty of similar
> discussions in the past in either ASF or ASF project that went the same way
> - not happy with simple system, implementing more complex one that had
> different "fairness" properties - and well we can pick one for
> multi-voting options if we want, I would not encourage us figuring out our
> own "approach" but simply pick from existing ones.
>
> ASF has a project called Apache Steve [2] that implements multiple voting
> methods that ASF recommends. You can read about those methods at STeVe
> page. We use it for board elections and we used it (and Single
> Transferable Vote [3] method) this year to select the new logo candidates.
> This is a "multi-winner" system - i.e. one that allows several winners.
> There is a single-winner variant of it - Instant Runoff Voting [4] that can
> be used instead - where you put your options in the sequence of preferences.
> Generally - the Single Transferable Vote is widely considered in ASF as
> the most "fair" and we use it for almost everything where multiple options
> are possible.
>
> Note that we do not have to use it of course (though we could ask infra to
> set it up for us) - this system is designed in the way that people have
> one-time ballots and that the voting is anonymous and can't be really
> traced to a person - and I think we are nowhere near that expectation. But
> we can use one of the methods it implements. In simple voting like ours, it
> would be rather easy to implement it manually in an shared spreadsheet (or
> we can find a ready-solution that is less complex than STeVe)
>
> *My proposal is not to invent our own - but to use Instant Runoff Voting
> (as the variant of the Apache-recommended STV).*
>
> More details on the Instant Runoff voting:
>
> People instead of voting +1, 0, -1 simply order the votes in preference.
> Then one or more eliminations are used to simulate multiple runoff
> elections. In each round, the candidate with the fewest first-preference
> votes (among the remaining candidates) is eliminated. This continues until
> only one candidate is left.
>
> [1] Electoral system: https://en.wikipedia.org/wiki/Electoral_system
> [2] ASF Steve project: https://steve.apache.org/
> [3] Single Transferable Vote
> https://en.wikipedia.org/wiki/Single_transferable_vote
> [4] Instant Runoff-voting
> https://en.wikipedia.org/wiki/Instant-runoff_voting
>
> J.
>
>
> On Thu, Oct 23, 2025 at 8:49 AM Amogh Desai <am...@apache.org> wrote:
>
>> I think I agree mostly with what Daniel says.
>>
>> We should arrive at a preference based voting to eliminate any sort of
>> voting power bias.
>>
>> I would second Daniel to have a rank based voting for ballots which can
>> have multiple choice: https://en.wikipedia.org/wiki/Instant-runoff_voting
>> .
>>
>> The ad-hoc +/- voting has more problems than we can imagine, and those
>> can be solved by ranked ballots -- giving everyone equal power while also
>> being easy to explain and implement.
>>
>> Thanks & Regards,
>> Amogh Desai
>>
>>
>> On Thu, Oct 23, 2025 at 9:20 AM Kiruban Kamaraj <kirubankamaraj@gmail.com
>> >
>> wrote:
>>
>> > IMHO, if we don't let people vote -1, how are devs supposed to raise
>> > legitimate concerns about an option? If someone votes -1, they should
>> > explain why - and hopefully devs are only doing this when they have real
>> > concerns, not just to push their favorite choice. If you have a solid
>> > reason to oppose something, just vote -1. If you don't care about
>> another
>> > option either way, then don't vote on it.
>> >
>> > I agree with what Jerek said.
>> >
>> > *it's not "who wins" but "which option wins". I don't absolutely care
>> who*
>> > *"wins" here, but which option has the most support.*
>> >
>> > On Thu, Oct 23, 2025 at 2:55 AM Daniel Standish via dev <
>> > dev@airflow.apache.org> wrote:
>> >
>> > > The DAG terminology vote I think has surfaced a problem with our
>> multiple
>> > > choice voting procedure.
>> > >
>> > > If you allow people to vote for multiple options, they seem to tend to
>> > use
>> > > it in a manner to signify their ranked preference. However, this
>> could
>> > > easily result in an option that doesn't have majority preference
>> getting
>> > > the win.
>> > >
>> > > E.g. suppose 4 people vote for option A, and 5 people vote for option
>> B
>> > +1
>> > > but also +0.5 for A. Then option A will win even though people prefer
>> > > option B 5 to 4.
>> > >
>> > > This is a bad outcome.
>> > >
>> > > It gets even stranger if you allow negative votes. Then you end
>> > > essentially invalidating other peoples votes, unless *everyone*
>> minuses
>> > all
>> > > of the options they don't favor. And even if everyone does that, then
>> > it's
>> > > hard to see how that gets to the outcome favored by most.
>> > >
>> > > With ranked choice voting, everyone votes for their most favored
>> choice,
>> > > but they can also rank all the options. If their most favored option
>> > does
>> > > not win, then their vote goes to their second favored option, and so
>> on.
>> > >
>> > > This is a better way to do this.
>> > >
>> > > I propose that when doing multiple choice votes, we do ranked choice,
>> > > instead of allowing people to just vote for multiple options with
>> plus or
>> > > minus votes.
>> > >
>> >
>>
>
Re: [DISCUSS] multiple choice voting procedure
Posted by Wei Lee <we...@gmail.com>.
> Otherwise that just encourages you to always vote multiple options
IMO, this is what we should do. We should review all the options and share our thoughts on each one. If we choose not to vote on a particular option, it can be interpreted as a +0, (I don't feel strongly about it, but I'm okay with this.) Our goal is not just to identify our favorite option, but to express our thoughts on each option presented.
> E.g. suppose 4 people vote for option A, and 5 people vote for option B +1
> but also +0.5 for A. Then option A will win even though people prefer
> option B 5 to 4.
>
> This is a bad outcome.
I don’t necessarily think this is a bad outcome. Those 5 people also believe A is a good idea, just not as good as B.
Best,
Wei
[1] https://www.apache.org/foundation/voting#expressing-votes-1-0-1-and-fractions
> On Oct 23, 2025, at 3:40 PM, Jarek Potiuk <ja...@potiuk.com> wrote:
>
> If we are not happy with the current system that disregards voting power
> and simply states for preferences on options - then yes we can change it .
> Actually this is an important part of governance - to change the rules if
> we are unhappy with them.
>
> And Actually we don't have to design our own system. ASF has us
> covered.Voting types are a very well researched subject and there are
> multiple voting systems we can choose from see "Electoral systems": [1] -
> each of them with different properties. There were plenty of similar
> discussions in the past in either ASF or ASF project that went the same way
> - not happy with simple system, implementing more complex one that had
> different "fairness" properties - and well we can pick one for
> multi-voting options if we want, I would not encourage us figuring out our
> own "approach" but simply pick from existing ones.
>
> ASF has a project called Apache Steve [2] that implements multiple voting
> methods that ASF recommends. You can read about those methods at STeVe
> page. We use it for board elections and we used it (and Single
> Transferable Vote [3] method) this year to select the new logo candidates.
> This is a "multi-winner" system - i.e. one that allows several winners.
> There is a single-winner variant of it - Instant Runoff Voting [4] that can
> be used instead - where you put your options in the sequence of preferences.
> Generally - the Single Transferable Vote is widely considered in ASF as the
> most "fair" and we use it for almost everything where multiple options are
> possible.
>
> Note that we do not have to use it of course (though we could ask infra to
> set it up for us) - this system is designed in the way that people have
> one-time ballots and that the voting is anonymous and can't be really
> traced to a person - and I think we are nowhere near that expectation. But
> we can use one of the methods it implements. In simple voting like ours, it
> would be rather easy to implement it manually in an shared spreadsheet (or
> we can find a ready-solution that is less complex than STeVe)
>
> *My proposal is not to invent our own - but to use Instant Runoff Voting
> (as the variant of the Apache-recommended STV).*
>
> More details on the Instant Runoff voting:
>
> People instead of voting +1, 0, -1 simply order the votes in preference.
> Then one or more eliminations are used to simulate multiple runoff
> elections. In each round, the candidate with the fewest first-preference
> votes (among the remaining candidates) is eliminated. This continues until
> only one candidate is left.
>
> [1] Electoral system: https://en.wikipedia.org/wiki/Electoral_system
> [2] ASF Steve project: https://steve.apache.org/
> [3] Single Transferable Vote
> https://en.wikipedia.org/wiki/Single_transferable_vote
> [4] Instant Runoff-voting
> https://en.wikipedia.org/wiki/Instant-runoff_voting
>
> J.
>
>
> On Thu, Oct 23, 2025 at 8:49 AM Amogh Desai <am...@apache.org> wrote:
>
>> I think I agree mostly with what Daniel says.
>>
>> We should arrive at a preference based voting to eliminate any sort of
>> voting power bias.
>>
>> I would second Daniel to have a rank based voting for ballots which can
>> have multiple choice: https://en.wikipedia.org/wiki/Instant-runoff_voting.
>>
>> The ad-hoc +/- voting has more problems than we can imagine, and those
>> can be solved by ranked ballots -- giving everyone equal power while also
>> being easy to explain and implement.
>>
>> Thanks & Regards,
>> Amogh Desai
>>
>>
>> On Thu, Oct 23, 2025 at 9:20 AM Kiruban Kamaraj <ki...@gmail.com>
>> wrote:
>>
>>> IMHO, if we don't let people vote -1, how are devs supposed to raise
>>> legitimate concerns about an option? If someone votes -1, they should
>>> explain why - and hopefully devs are only doing this when they have real
>>> concerns, not just to push their favorite choice. If you have a solid
>>> reason to oppose something, just vote -1. If you don't care about another
>>> option either way, then don't vote on it.
>>>
>>> I agree with what Jerek said.
>>>
>>> *it's not "who wins" but "which option wins". I don't absolutely care
>> who*
>>> *"wins" here, but which option has the most support.*
>>>
>>> On Thu, Oct 23, 2025 at 2:55 AM Daniel Standish via dev <
>>> dev@airflow.apache.org> wrote:
>>>
>>>> The DAG terminology vote I think has surfaced a problem with our
>> multiple
>>>> choice voting procedure.
>>>>
>>>> If you allow people to vote for multiple options, they seem to tend to
>>> use
>>>> it in a manner to signify their ranked preference. However, this could
>>>> easily result in an option that doesn't have majority preference
>> getting
>>>> the win.
>>>>
>>>> E.g. suppose 4 people vote for option A, and 5 people vote for option B
>>> +1
>>>> but also +0.5 for A. Then option A will win even though people prefer
>>>> option B 5 to 4.
>>>>
>>>> This is a bad outcome.
>>>>
>>>> It gets even stranger if you allow negative votes. Then you end
>>>> essentially invalidating other peoples votes, unless *everyone* minuses
>>> all
>>>> of the options they don't favor. And even if everyone does that, then
>>> it's
>>>> hard to see how that gets to the outcome favored by most.
>>>>
>>>> With ranked choice voting, everyone votes for their most favored
>> choice,
>>>> but they can also rank all the options. If their most favored option
>>> does
>>>> not win, then their vote goes to their second favored option, and so
>> on.
>>>>
>>>> This is a better way to do this.
>>>>
>>>> I propose that when doing multiple choice votes, we do ranked choice,
>>>> instead of allowing people to just vote for multiple options with plus
>> or
>>>> minus votes.
>>>>
>>>
>>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@airflow.apache.org
For additional commands, e-mail: dev-help@airflow.apache.org
Re: [DISCUSS] multiple choice voting procedure
Posted by Jarek Potiuk <ja...@potiuk.com>.
If we are not happy with the current system that disregards voting power
and simply states for preferences on options - then yes we can change it .
Actually this is an important part of governance - to change the rules if
we are unhappy with them.
And Actually we don't have to design our own system. ASF has us
covered.Voting types are a very well researched subject and there are
multiple voting systems we can choose from see "Electoral systems": [1] -
each of them with different properties. There were plenty of similar
discussions in the past in either ASF or ASF project that went the same way
- not happy with simple system, implementing more complex one that had
different "fairness" properties - and well we can pick one for
multi-voting options if we want, I would not encourage us figuring out our
own "approach" but simply pick from existing ones.
ASF has a project called Apache Steve [2] that implements multiple voting
methods that ASF recommends. You can read about those methods at STeVe
page. We use it for board elections and we used it (and Single
Transferable Vote [3] method) this year to select the new logo candidates.
This is a "multi-winner" system - i.e. one that allows several winners.
There is a single-winner variant of it - Instant Runoff Voting [4] that can
be used instead - where you put your options in the sequence of preferences.
Generally - the Single Transferable Vote is widely considered in ASF as the
most "fair" and we use it for almost everything where multiple options are
possible.
Note that we do not have to use it of course (though we could ask infra to
set it up for us) - this system is designed in the way that people have
one-time ballots and that the voting is anonymous and can't be really
traced to a person - and I think we are nowhere near that expectation. But
we can use one of the methods it implements. In simple voting like ours, it
would be rather easy to implement it manually in an shared spreadsheet (or
we can find a ready-solution that is less complex than STeVe)
*My proposal is not to invent our own - but to use Instant Runoff Voting
(as the variant of the Apache-recommended STV).*
More details on the Instant Runoff voting:
People instead of voting +1, 0, -1 simply order the votes in preference.
Then one or more eliminations are used to simulate multiple runoff
elections. In each round, the candidate with the fewest first-preference
votes (among the remaining candidates) is eliminated. This continues until
only one candidate is left.
[1] Electoral system: https://en.wikipedia.org/wiki/Electoral_system
[2] ASF Steve project: https://steve.apache.org/
[3] Single Transferable Vote
https://en.wikipedia.org/wiki/Single_transferable_vote
[4] Instant Runoff-voting
https://en.wikipedia.org/wiki/Instant-runoff_voting
J.
On Thu, Oct 23, 2025 at 8:49 AM Amogh Desai <am...@apache.org> wrote:
> I think I agree mostly with what Daniel says.
>
> We should arrive at a preference based voting to eliminate any sort of
> voting power bias.
>
> I would second Daniel to have a rank based voting for ballots which can
> have multiple choice: https://en.wikipedia.org/wiki/Instant-runoff_voting.
>
> The ad-hoc +/- voting has more problems than we can imagine, and those
> can be solved by ranked ballots -- giving everyone equal power while also
> being easy to explain and implement.
>
> Thanks & Regards,
> Amogh Desai
>
>
> On Thu, Oct 23, 2025 at 9:20 AM Kiruban Kamaraj <ki...@gmail.com>
> wrote:
>
> > IMHO, if we don't let people vote -1, how are devs supposed to raise
> > legitimate concerns about an option? If someone votes -1, they should
> > explain why - and hopefully devs are only doing this when they have real
> > concerns, not just to push their favorite choice. If you have a solid
> > reason to oppose something, just vote -1. If you don't care about another
> > option either way, then don't vote on it.
> >
> > I agree with what Jerek said.
> >
> > *it's not "who wins" but "which option wins". I don't absolutely care
> who*
> > *"wins" here, but which option has the most support.*
> >
> > On Thu, Oct 23, 2025 at 2:55 AM Daniel Standish via dev <
> > dev@airflow.apache.org> wrote:
> >
> > > The DAG terminology vote I think has surfaced a problem with our
> multiple
> > > choice voting procedure.
> > >
> > > If you allow people to vote for multiple options, they seem to tend to
> > use
> > > it in a manner to signify their ranked preference. However, this could
> > > easily result in an option that doesn't have majority preference
> getting
> > > the win.
> > >
> > > E.g. suppose 4 people vote for option A, and 5 people vote for option B
> > +1
> > > but also +0.5 for A. Then option A will win even though people prefer
> > > option B 5 to 4.
> > >
> > > This is a bad outcome.
> > >
> > > It gets even stranger if you allow negative votes. Then you end
> > > essentially invalidating other peoples votes, unless *everyone* minuses
> > all
> > > of the options they don't favor. And even if everyone does that, then
> > it's
> > > hard to see how that gets to the outcome favored by most.
> > >
> > > With ranked choice voting, everyone votes for their most favored
> choice,
> > > but they can also rank all the options. If their most favored option
> > does
> > > not win, then their vote goes to their second favored option, and so
> on.
> > >
> > > This is a better way to do this.
> > >
> > > I propose that when doing multiple choice votes, we do ranked choice,
> > > instead of allowing people to just vote for multiple options with plus
> or
> > > minus votes.
> > >
> >
>
Re: [DISCUSS] multiple choice voting procedure
Posted by Amogh Desai <am...@apache.org>.
I think I agree mostly with what Daniel says.
We should arrive at a preference based voting to eliminate any sort of
voting power bias.
I would second Daniel to have a rank based voting for ballots which can
have multiple choice: https://en.wikipedia.org/wiki/Instant-runoff_voting.
The ad-hoc +/- voting has more problems than we can imagine, and those
can be solved by ranked ballots -- giving everyone equal power while also
being easy to explain and implement.
Thanks & Regards,
Amogh Desai
On Thu, Oct 23, 2025 at 9:20 AM Kiruban Kamaraj <ki...@gmail.com>
wrote:
> IMHO, if we don't let people vote -1, how are devs supposed to raise
> legitimate concerns about an option? If someone votes -1, they should
> explain why - and hopefully devs are only doing this when they have real
> concerns, not just to push their favorite choice. If you have a solid
> reason to oppose something, just vote -1. If you don't care about another
> option either way, then don't vote on it.
>
> I agree with what Jerek said.
>
> *it's not "who wins" but "which option wins". I don't absolutely care who*
> *"wins" here, but which option has the most support.*
>
> On Thu, Oct 23, 2025 at 2:55 AM Daniel Standish via dev <
> dev@airflow.apache.org> wrote:
>
> > The DAG terminology vote I think has surfaced a problem with our multiple
> > choice voting procedure.
> >
> > If you allow people to vote for multiple options, they seem to tend to
> use
> > it in a manner to signify their ranked preference. However, this could
> > easily result in an option that doesn't have majority preference getting
> > the win.
> >
> > E.g. suppose 4 people vote for option A, and 5 people vote for option B
> +1
> > but also +0.5 for A. Then option A will win even though people prefer
> > option B 5 to 4.
> >
> > This is a bad outcome.
> >
> > It gets even stranger if you allow negative votes. Then you end
> > essentially invalidating other peoples votes, unless *everyone* minuses
> all
> > of the options they don't favor. And even if everyone does that, then
> it's
> > hard to see how that gets to the outcome favored by most.
> >
> > With ranked choice voting, everyone votes for their most favored choice,
> > but they can also rank all the options. If their most favored option
> does
> > not win, then their vote goes to their second favored option, and so on.
> >
> > This is a better way to do this.
> >
> > I propose that when doing multiple choice votes, we do ranked choice,
> > instead of allowing people to just vote for multiple options with plus or
> > minus votes.
> >
>
Re: [DISCUSS] multiple choice voting procedure
Posted by Kiruban Kamaraj <ki...@gmail.com>.
IMHO, if we don't let people vote -1, how are devs supposed to raise
legitimate concerns about an option? If someone votes -1, they should
explain why - and hopefully devs are only doing this when they have real
concerns, not just to push their favorite choice. If you have a solid
reason to oppose something, just vote -1. If you don't care about another
option either way, then don't vote on it.
I agree with what Jerek said.
*it's not "who wins" but "which option wins". I don't absolutely care who*
*"wins" here, but which option has the most support.*
On Thu, Oct 23, 2025 at 2:55 AM Daniel Standish via dev <
dev@airflow.apache.org> wrote:
> The DAG terminology vote I think has surfaced a problem with our multiple
> choice voting procedure.
>
> If you allow people to vote for multiple options, they seem to tend to use
> it in a manner to signify their ranked preference. However, this could
> easily result in an option that doesn't have majority preference getting
> the win.
>
> E.g. suppose 4 people vote for option A, and 5 people vote for option B +1
> but also +0.5 for A. Then option A will win even though people prefer
> option B 5 to 4.
>
> This is a bad outcome.
>
> It gets even stranger if you allow negative votes. Then you end
> essentially invalidating other peoples votes, unless *everyone* minuses all
> of the options they don't favor. And even if everyone does that, then it's
> hard to see how that gets to the outcome favored by most.
>
> With ranked choice voting, everyone votes for their most favored choice,
> but they can also rank all the options. If their most favored option does
> not win, then their vote goes to their second favored option, and so on.
>
> This is a better way to do this.
>
> I propose that when doing multiple choice votes, we do ranked choice,
> instead of allowing people to just vote for multiple options with plus or
> minus votes.
>
Re: [DISCUSS] multiple choice voting procedure
Posted by Pierre Jeambrun <pi...@gmail.com>.
Also worth considering in the current state, someone can vote multiple +1
+0,5 ec..., virtually 'increasing' their voting power compared to people
that only +1 a single option.
(Sum of your votes should be +1 or something like that to balance out
values).
Otherwise that just encourages you to always vote multiple options so you
get your 'second' favorite option an edge.
On Wed, Oct 22, 2025 at 11:26 PM Daniel Standish via dev <
dev@airflow.apache.org> wrote:
> The DAG terminology vote I think has surfaced a problem with our multiple
> choice voting procedure.
>
> If you allow people to vote for multiple options, they seem to tend to use
> it in a manner to signify their ranked preference. However, this could
> easily result in an option that doesn't have majority preference getting
> the win.
>
> E.g. suppose 4 people vote for option A, and 5 people vote for option B +1
> but also +0.5 for A. Then option A will win even though people prefer
> option B 5 to 4.
>
> This is a bad outcome.
>
> It gets even stranger if you allow negative votes. Then you end
> essentially invalidating other peoples votes, unless *everyone* minuses all
> of the options they don't favor. And even if everyone does that, then it's
> hard to see how that gets to the outcome favored by most.
>
> With ranked choice voting, everyone votes for their most favored choice,
> but they can also rank all the options. If their most favored option does
> not win, then their vote goes to their second favored option, and so on.
>
> This is a better way to do this.
>
> I propose that when doing multiple choice votes, we do ranked choice,
> instead of allowing people to just vote for multiple options with plus or
> minus votes.
>