Re: [gentoo-project] [RFC] Triumvirate in Gentoo

[Alec Warner]

On Thu, Jun 4, 2020 at 12:30 AM Michał Górny <mgorny@g.o> wrote:

> Hello, everyone.
>
> This is something I wanted to discuss back in April but due to the peak
> of covid pandemic I've delayed it.  Today things seem to be improving
> a bit, at least in Europe, so I'd like to bring it up now, especially
> with the elections coming soon.
>
>
> Gentoo is technically led by two bodies -- the Council and the Trustees.
> While this somewhat works for many years, people have repeatedly pointed
> out that it's far from perfect and that it is preventing Gentoo from
> gaining more popularity.  Some of them are looking into the times of
> BDFL with longing, others are considering it the worst thing ever.
> Nevertheless, there are problems with the current state of things.
>
>
Is Gentoo being popular a goal? How does the current structure prevent
anything?

I think the existing leadership is not very proactive (in both bodies.)
Often the leadership is focussed on adjudication (someone raises an issue,
leaders consider and issue a judgement, issue is resolved / shelved.) I
don't see either body putting forth a vision for Gentoo, nor leading in
what I'd expect would be a more top-down fashion (perhaps necessary to
implement any vision.) I'm curious to hear more about this; is this lack of
vision a lack of remit? Or is it just too hard? Or no one is interested in
a vision?


> Firstly, we have two leading bodies and still no clear distinction
> between their roles.  Some developers agree on split being here, some
> developers put it elsewhere but in the end, nothing has been really
> decided.  From time to time one of the bodies tries to push their border
> forward, then backs down and we're back where we started.
>
> Secondly, for historical reasons the both bodies are elected by two
> electorates that only partially overlap.  Surely, today the overlap is
> reasonable but is there any real reason for different people to elect
> both bodies?  In the end, it is entirely possible for one body to
> arbitrarily change their electorate and made it completely disjoint.
>
> Thirdly, large governing bodies don't really work.  Instead of having
> one consistent vision of Gentoo, we have 12.  What we get is a semi-
> random combination of parts of their visions that just happened to hit
> majority in their votes.  It gets absurd to the point that a body can
> make half-way decisions just because first half passed vote
> and the second didn't (remember closing -dev but leaving -project
> open?).
>

I seem to recall that decision being undone by a later council, so I think
the right thing happened in the end ;)


> Compromises are sometimes good and sometimes horrible.  If one dev wants
> to paint the bikeshed red and another one blue, mixings the two colors
> doesn't really get either what he wants.  You just get a third color
> that nobody is happy with, and in the best case you could say that
> neither of them got what he wanted.
>

"A really good compromise is the one that leaves both sides equally
dissatisfied." (see below)


>
> BDFL is not a perfect solution either.  While having one has the obvious
> advantage of having a single consistent vision for the distribution,
> giving absolute power to a single person creates a fair risk of abuse.
> This is not something most of Gentoo devs would agree to.
>

I don't think a BFDL is a necessary condition for a vision. Let's have a
hypothetical where in the existing system someone proposes a broad vision
for Gentoo and its ratified by all leadership bodies. What effect might
that have on developers whose work is not compatible? Would they no longer
contribute; and is that a thing that we want? Part of the benefit of the
compromise model is that you compromise because both sides have something
to offer and you want to come to some mutually beneficial arrangement.
Typically in Gentoo it means we have fairly loose development to facilitate
a broad base of developers. If we tighten up our development process we may
lose people who are not on board (because the compromise is no longer
facilitating their contributions.)


>
>
> All that said, I'd propose to meet in the middle -- following
> the ancient tradition, establish a triumvirate in Gentoo.  It would be:
>
> 1. Technical lead -- a person with exceptional technical talents that
> would build the vision of Gentoo from technical perspective, i.e. make
> a distribution that people would love using.  Initially, this role could
> be taken by the QA lead.
>
> 2. Social lead -- a person with exceptional social skills that would
> build the vision of Gentoo from community perspective, i.e. make
> a distribution that people would love contributing to.  Initially, this
> role would taken by the ComRel lead.
>
> 3. Organization lead -- a person with (exceptional) business skills that
> would take care of all the financial and organizational aspects of
> Gentoo, i.e. make a distribution that sustains.  Initially, this role
> would be taken by the Foundation president.
>
> Three seems to be a very good number -- on one hand, it's more than one,
> so the others can stop any single one from getting absolute power.
> On the other, it's small enough for them to be able to actively work
> together and directly establish a common set of goals (i.e. via
> an agreement rather than a majority vote).
>

I think you still need that 3 member triumvirate to practice a *lot* of
delegation in order to lead a 100+ person project. This is something we
could do a lot better at.

-A


>
>
> WDYT?
>
> --
> Best regards,
> Michał Górny
>
>