Adjust AC appeal vote threshold based on participation

[frivoal]

(Split from #882).

To avoid making momentous decisions based on a slim majority in vote with low participation, the W3C's bylaws have provisions for what they call "Requisite Member Vote" where decisions are confirmed by :

• a 75% majority when participation is at or below 15%
• a 2/3rds majority when participation is between 15% and 20%
• a 50% majority when participation is at or above 20%

I think it would be reasonable to adopt the same concept for AC appeals, which currently can overturn decisions via a 50% majority without quorum or any other consideration for the level of participation.

[frivoal]

This would look something like this: 7bc5058

[frivoal]

From the AB:

RESOLUTION: Adopt supermajority requirements for low-quorum AC Appeals using same ramp as Bylaws; add minimum quorum of 5%.

[martinthomson]

I disagree with this decision. It's needlessly complicated and it results in non-linearity that encourages odd voting patterns.

[martinthomson]

Having looked at the bylaws more closely, I want to revise my statement above.

Why would this not instead refer to the bylaws. The change in #901 sets up different thresholds, where it seems like the rationale was to have the rules be congruent. A reference to the "Requisite Member Vote" rules in the bylaws (which fail to set a minimum threshold, so a single member could affirm a decision) would also benefit from updates to that definition.

[frivoal]

This issue has been closed on the basis of an AB resolution. The Process CG follows AB instructions when they exist (and refers to the AB for resolution of any non-consensual situation), so if you would like this question to be revisited, you need to ask the AB. https://github.com/w3c/AB-memberonly/issues would be the right place.

With that said, here are some further thoughts about the substance of your comment:

it results in non-linearity that encourages odd voting patterns.

I agree it is non-linear / discontinuous. (I too would have preferred a continuous function, but mathematical elegance was not he main criterion followed by the laywers who drafted this concept for the bylaws.)

I disagree that this is exploitable and encourages odd/strategic voting. Those voting against a proposal might feel like they should refrain from voting if (a) the vote is currently being rejected (b) we are nearing the vote count that would take us beyond the next threshold and which would lower the proportions of yes votes needed to make the vote pass, and (c) they could be confident that others will refrain from voting too, so that no one else pushes us over the threshold. But these 3 things aren't knowable, so voters cannot adjust their behavior accordingly even if they wanted to (in particular, even if (a) and (b) could be knowable in an election where voters are given a high level of transparency onto ongoing votes, they still cannot know how many others will vote after them).

Why would this not instead refer to the bylaws

Maintaining minimum coupling between The bylaws and the Process has been a deliberate decision from the very first iteration of the bylaws, for a variety of reasons, including:

• Consortium Member in the bylaws sense and Member in the Process sense are very similar concepts, but not guaranteed to be 100% identical, in part due to termination clauses, and clauses about Member Associations. The fact that they're not identical may be regrettable, but it is what it is.
• Even if they are identical, Consortium Member Representative, and AC Rep are not required to be the same person.
• The evolution of the Process and the evolution of the bylaws are governed by different rules. Normative cross references between the two would make that challenging.
• The Process predates the bylaws, and is independent of them. Maintaining that independence also gives us more flexibility in case we ever want/need to evolve the legal form of the W3C corporation into something different.

All in all, even if we want similarities to keep things simple, avoiding normative dependencies between the two as much as possible is a deliberate choice, and a good one I think.