PEP 8014 – The Commons Governance Model
- Jack Jansen
This PEP proposes a governance model with as few procedures, defined terms and percentages as possible. It may also be called The Anarchist Governance Model but uses Commons for now because of possible negative connotations of the term Anarchist to some audiences.
The basic idea is that all decisions are in principle voted on by the whole community, but in practice voted on by only a subset of the community. A subset, because although the whole community is entitled to vote in practice it will always be only a small subset that vote on a specific decision. The vote is overseen by an impartial council that judges whether the decision has passed or not. The intention is that this council bases its decision not only on the ratio of yes and no votes but also on the total number of votes, on the gravity of the proposal being voted on and possibly the individual voters and how they voted. Thereby this council becomes responsible for ensuring that each individual decision is carried by a sufficient majority.
PEP 8014 was rejected by a core developer vote described in PEP 8001 on Monday, December 17, 2018.
PEP 8016 and the governance model it describes were chosen instead.
The Commons Governance Model tries to ensure that all decisions are endorsed by, or at least is acceptable to, a sufficient majority of the Python community.
Unfortunately the previous paragraph has two terms that are very hard to quantify in the general case: sufficient majority and Python community. This is because both terms in reality depend on the specific case that is being decided. To give an example of this difficulty: for a PEP that proposes a backward-compatible change to some API a simple majority of the core developers that were interested in voting on the PEP in the first place is probably sufficient. But for a change that has more farreaching consequences such as a Python3 to Python4 transition a real majority may be wanted, and a demonstration that at least there seems to be sufficient support in the user base. And for a change that transcends the Python-the-language, such as decisions on abolishing non-inclusive language, it becomes very vague.
The Commons Governance Model attempts to sidestep this issue by not defining what the terms sufficient majority and Python community mean in the general case, by proposing a body that will decide so in specific cases.
The model proposes creating a Council of Elders that oversees the decision process, determining whether a specific proposal has enough support on a case-by-case basis. There will be a vote on every individual PEP, and the Council of Elders will declare whether the outcome of the vote is sufficient to carry the decision in this specific case.
The model addresses only the roles traditionally held by the BDFL in the decision process, not other roles.
The term Commons in the model name is loosely based on its historic use as a shared resource to be used by all and cared for by all. The picture you should have in mind with this model is a sizeable group of peasants discussing some plan for the future on the village green on a warm summer evening, after which the vote is taken and the village elders pronounce the outcome. Then the banquet begins.
The Commons Governance Model is different from most of the other governance proposals (with the possible exception of 8012), because it explicitly places supreme power with the whole community.
The rationale for the model is that a model that casts everything in concrete will have unintended negative side effects. For example, a governance model that assigns voting rights to Python committers may cause an individual not to be accepted as a committer because there are already a lot of committers from the company the new candidate works for.
As another example, setting a fixed percentage for PEP acceptance may lead to party-formation amongst the voters and individual PEPs no longer be being judged on individual merit but along party lines (if you support my PEP I will support yours).
There is also the issue that one-person-one-vote is not the best model for something like Python. Again an example: in case of a split vote (or a vote sufficiently close to being split) the opinion of core developer Guido van Rossum should probably outweigh the opinion of core developer Jack Jansen. Trying to formalize this in a voting model is going to lead to a very complex model, that is going to be wrong on boundary cases anyway. The model presented here leaves deciding on such issues to the (hopefully sensible) council of elders.
All important decisions go through a PEP process. Each PEP has someone responsible for it, called the author here, but that does not have to be a single person, and it does not have to be the person that actually wrote the text. So for author you could also read champion or shepherd or something like that.
The PEP author is responsible for organizing a vote on the PEP. This vote is public, i.e. the voters are identified and the results are known to all. Voting may be simple +1/0/-1, but might also be extended with +2/-2 with a very terse explanation why the voter feels very strong about the issue. Such an annotation would serve as an explanation to the Council of Elders. Voters are annotated with their community status (core developer, etc).
The vote is clearly separated from the discussion, by using a well-defined Discourse category or tag, a special mailing list or a similar technical method (such as a website vote.python.org where people have to log in so their community status can be automatically added, and their identity can be somewhat confirmed).
The PEP author presents the PEP and the vote results to the Council of Elders. The council ponders two things:
- the PEP gravity and its implications,
- the measureable vote results (how many people voted, which individuals voted, what they voted).
They pronounce a tentative decision on whether the vote passed and this decision is published.
If the decision is that the vote results do not demonstrate enough support from the community for the decision the burden is on the author to try and gather more support and resubmit the vote at a later date. Alternatively the author can retract the proposal. The period for gathering more support is time-limited, a month seems a reasonable time, if no vote has been resubmitted after that period the proposal is rejected.
If the tentative decision is that the results do demonstrate enough support a fairly short waiting period starts (in the order of weeks). During this period anyone can appeal to the Council of Elders, but only on the grounds that the vote does not reflect a sufficient majority of the community. After the waiting period the council pronounces a final decision. The PEP is either accepted or, if the council is swayed by an appeal, goes back to the state where more support has to be demonstrated.
Council of Elders
The intention of the Council of Elders is that they, together, are capable of judging whether the will of the Python community is upheld in a specific vote.
The Council of Elders is not a replacement of the BDFL by a group of people with the same power as the BDFL: it will not provide guidance on the direction of Python, it only attempts to ensure the outcome of a vote represents the will of the community.
The Council of Elders is not like the US Supreme Court, which has actual decision power, the council only oversees the voting process to ensure that the community is represented in the vote. And the Council of Elders is most definitely not like the Spanish Inquisition, because fear, surprise and ruthless efficiency are things we can do without (but there is some merit in using the cute scarlet regalia).
The council is somewhat like the Dutch Hoge Raad (which is unfortunately often translated as Supreme Court in English) in that they judge the process and the procedures followed and can only send cases back for a renewed judgement.
It is also somewhat like the election commission that many countries have (under different names) in that it oversees elections.
The council members are volunteers, and most likely have other roles within the Python community as well (not to mention a life outside Python). This means that the workload on the members should be kept to a minimum. It also means that it should be clear when an individual council members speak as council member and when they speak as themselves. And we should care about the emotional load: council members should not be held accountable for decisions by random flamers on the Python mailing list.
The proposal attempts to minimize the workload through two methods:
- Most of the actual work is to be done by the PEP author and the community, the Council of Elders does not organize the vote and tally the results.
- The idea behind the first tentative decision is mistakes by the Council
of elders (misjudging how far-reaching a PEP is, most likely) are not fatal, because
the community has a chance to point out these mistakes.
Practically speaking this means that the tentative decision can be taken by a subset of the council, depending on the community to correct them. Getting seven hard-working professionals together every two weeks, even by email, may be a bit much to ask.
Clarifying when an individual Elder speaks on behalf of the Council is probably best done by using a special email address, or some Discourse topic into which only Elders can post. There is an analogy here with the Pope speaking Ex Cathedra or just as himself (in which case he is not infallible). The elders are most likely respected members of the community and it would be a bad idea if they feel they cannot voice their personal opinion on a PEP because they are on the council.
Discussion of community members with the Council of Elders, i.e. when appealing a decision, should be done in a different forum (Discourse topic, mailing list).
The decisions of the Council of Elders should be seen as decisions of the council as a whole, not as decisions of the individual members. In a first implementation Elders should post under their own name (with the fact that they speak as a council member conferred by the topic they post to, or possibly a special badge). If it turns out that Elders become individual targets for ad-hominem attacks we should revisit this and come up with some method of anonymity.
Limitation of freedom
If a specific vote has a true majority (for or against) of core team members (more than 50% + 1 of all core team members) that outcome passes. If a specific vote has a true majority (for or against) of PSF voting members (more than 50% + 1) that outcome passes. And, for completeness, if both of the previous statements are true but with opposite outcomes the core team members win.
The main reason for having this limitation is that it allows decisions to be made (albeit with effort) if there is no functioning Council of Elders at any particular moment.
The council should not be too big nor too small, probably somewhere between 5 and 10 members. There is no reason to fix this number. The members should be knowledgeable about Python and the Python community, and willing to be impartial while operating as part of the council. Council members may be core developers but this is not a requirement.
Everyone in the community should feel represented by the council so it would be good if the council is diverse:
- scientists and technologists,
- progressives and conservatives (with respect to the Python language),
- people with different cultural backgrounds, genders, age,
But: this should hold for the council as a whole. Individual council members should not be seen as representing a specific interest group.
Because the powers of the council are purely procedural it is probably good if members serve for a fairly long time. However, it would still be good if the council was reinstated regularly. Therefore, the suggestion is to have the council operate under the PSF umbrella and be subject of a yearly vote of confidence. This vote is for the council as a whole: people who vote against the council should be aware that they are basically saying “Python is better off without a Council of Elders than with you lot”.
The council normally co-opts new Elders, probably because an individual is seen to have knowledge about a specific part of the Python community (or language) in which the council is lacking. Everyone is free to suggest new Elders to the council (including themselves) but the council is free to ignore the suggestion. Council members should be free to retire at any time. An individual council member can be retired by a unanimous vote by the rest of the council.
There is an emergency brake procedure to get rid of a non-functioning council. A single Elder or a group of 10 core developers or PSF voting members can ask for an immediate reinstating vote of the council as a whole (presumably with the intention that the council lose their mandate). If this vote has been requested by an Elder that individual immediately lose their council position, independent of the outcome of the vote. If the vote has been requested by community members and the council is reinstated this procedure cannot be invoked again for a year.
If there is no functioning council (the current initial situation, or after the council have lost their mandate after a vote of no confidence) an initial council must be selected. Through the normal communication channels (discourse, mailing lists) members can be suggested by anyone (including themselves). After discussion amongst the nominees and in the whole community a group of at least three individuals should emerge that ask for an initial vote to instate them as Council of Elders. The intention of this procedure is that by the time such a group of individuals emerges and asks for a vote of confidence they expect an overwhelming mandate.
This PEP does not handle other roles of the BDFL, only the voting process. Most importantly, the direction of Python in the long term is not expected to be handled by the Council of Elders. This falls to the community as a whole (or to individual members of the community, most likely).
There is also the role of figurehead or spokesperson to represent Python and the Python community to the outside world. Again, this is not a role that should be handled by the Council of Elders, in my opinion, but by some other person or body.
Note that this proposal most likely favors conservatism over progression. Or, at least, the danger of it leading to stagnation is bigger than the danger of it leading to reckless blazing ahead into unknown territories. So: we should realise that it is unlikely that a PEP like PEP 572 will pass if this model is in place.
This document has been placed in the public domain.
Last modified: 2022-01-21 11:03:51 GMT