[FFmpeg-devel] Democratization work in progress draft v2

[Soft Works]

> -----Original Message-----
> From: ffmpeg-devel <ffmpeg-devel-bounces at ffmpeg.org> On Behalf Of Leo
> Izen
> Sent: Sunday, February 2, 2025 3:26 AM
> To: ffmpeg-devel at ffmpeg.org
> Subject: Re: [FFmpeg-devel] Democratization work in progress draft v2
> 
> On 1/31/25 11:01 AM, Soft Works wrote:

[..]

Hi Leo,

> > How about a quadratic attenuation of past commit counts, so that
> older commits count less than more recent ones?

Shortly after sending I realized that it's exponential, not quadratic (which was the initial idea), sorry for the mixup.

> Such quadratic metrics tend to be such that for currently active
> contributors, it roughly correlates with square root of commit count
> (which is an increasing function) and therefore isn't meaningfully
> different.

I'm not sure whether I can follow. Do you mean for 

1. currently active contributors who have been active in the past
2. currently active contributors who have not been active in the past
3. for both equally?

If you mean that it's equal for all in (1), that would be just like intended as the goal would be to give them more voting power than those in (2).
It would also give those a vote who have contributed in the past but are no longer active, yet decreasing over time.


> A similar thing was discovered in the N-papers-cited-N-times-each
> metric
> which was popular at one point in academia as an alternative to
> citation
> count. It turned out to maximize the area of an axis-bounded square
> when
> contributions were plotted, which is why for naturally occurring data
> it
> correlated pretty well with the square root of total citation count.
> 
> While this isn't an entirely analogous situation, most contributors
> who
> are active have been active since they started contributing, 
> so this
> doesn't do a whole lot except to pick out people who used to be
> active
> and then stopped and then started up again.

I'm not sure how it was done in the case you are referring to, but it sounds like the time axis there would have been scaled  to the begin of their activity, while in this case, it would be absolute.

(please correct me if I'm misinterpreting)

Also, in the academic area, one usually doesn't stop activity like it happens in case of ffmpeg, for which one would be given decreasing weight of votes over time.


I'm not actually proposing this as a model, but some had said that contributions from past times (earlier than the 3-yr GA range) should give them more weight in voting, and my answer to that is, if this would be done, then it should be at least in a way that recent contributions will count more than older contributions.

What was the outcome in the academic world - back to citation count?

Best
sw