Managing Gentoo - a study in quotes [LWN.net]

People outside of the Gentoo Linux project may be surprised to learn that the Gentoo developers are currently electing a new management council. Unlike, say, Debian, Gentoo tends to do a fair amount of its deliberations out of public view. There has recently been a discussion, however, which has brought out some of the concerns that Gentoo developers have. Here are some excerpts.

I started my fourth year as a Gentoo developer in June, and Gentoo's changed a lot since I started back in 2003. We've become a drastically more democratic organization. But the question remains - _Is this a good thing?_

When I think about where Gentoo was when we turned into a democracy years ago, and where Gentoo is now, I don't see much of a difference on the large scale. We lack any global vision for where Gentoo is going, we can't agree on who our audience is, and everyone's just working on pretty much whatever they feel like. [...]

I'm not the only one to suggest that a democracy isn't the most productive way to run Gentoo. When people wanted to change in how Gentoo was run, democracy was the only option considered, rather than simply changing the leaders. There's an ongoing assumption that if problems exist, it must be somewhere in the structure rather than in the people.

If I could go back in time a couple of years and prevent this democracy from ever happening, I would. If I could fix these problems myself, I would. But it requires buy-in from the entire Gentoo community if we're to do anything about it.

-- Donnie Berkholz

In addition to the conclusion that too much freedom has entered the life-blood that drives Gentoo it is also often the case that from the stance of upper management there is not enough freedom given. Part of what paralyzes the Council and devrel and any other historical body that has tried to keep Gentoo healthy is that there is an understanding that they can only act as a whole...as individuals none of them have power as there is fear that a rogue person in a position to abuse their responsibility will do so. It is my contention that with a body of multiple individuals such as the Council that there would be the ability to recognize and mitigate the damage done by such a rogue. I'd posit that by voting someone onto the council you are saying that you trust them enough to carry this duty on their shoulders. The Council itself should not be just a technical body to validate the merits of GLERs and/or emerging projects, it (or some other yet to be established group) has to carry the solemn duty of carrying Gentoo into the future, nurturing it as only a parent could....

All in all I suppose that is the platform that I am running on for this years Council...take it for what you will but that is where I stand.

-- Daniel Ostrow

If there's a lack of respect at the moment, it's not for devrel. It's between individual developers, who either do not value each other as people, or do not value each other as contributors.

A good way to sort that out is to get them together in the physical world, and use group de-polarisation exercises to help folks understand that their view of the world isn't the only view that is valid. This is why I'm hoping to see Gentoo establish a regular international dev conference. You'll find that the vast majority of issues won't arise once folks actually know each other better - and the personality clashes that are left are easier to see for what they are.

-- Stuart Herbert

Maybe its a cultural thing between some of us, or maybe its the 'pre-daniel' versus 'post-daniel' devs. I'm curious the demographics of our active developers that were on prior to daniel's leaving compared to those who joined after. To most of the recent active folks, they never knew what it was like before. Hell, I just got on towards the tail end of the daniel-era, so I don't have much validity in that realm myself! But I do remember how it used to be and how well we did things and how we usually respected each other in some fashion or another.

I'm afraid those days are in the past unless some kind of fork happens where the folks who think we need a leader go their way and the folks who prefer the leader-by-committee approach go their way. We all hate forks, none of us have time for forks, but looking at the dividing line, I don't see how we'll be able to compromise with out adding more policies and BS.

-- Lance Albertson

It's very easy to claim that "there are too many flamewars", even if that isn't actually true. It's hard to claim "Portage needs replacing, the tree has huge QA issues, several archs are horribly unmaintained and too many developers don't have a clue what they're doing" because a) they're difficult problems to address, b) if you do say them, Condorcet ensures that you won't get elected and c) you might be expected to fix them.

Most of these problems could be solved if we had a council that was far less spineless, a council that's prepared to address the *real* issues rather than doing nothing, a council that shows leadership and provides direction where it's needed without screwing things up where it's not.

-- Ciaran McCreesh

I definitely agree here. What has made me decide to run for the council is my wish to see things improve before we honestly do start hemorrhaging developers. We have seen indications that it is coming, but it hasn't started quite yet. A strong leadership is needed to give us direction where needed, and also to leave people well enough alone where it is not needed.

-- Chris Gianelloni

At the top level, the council, in its present form does not manage Gentoo. It can't, it's pretty much disempowered as a management organisation due to the rules for its agenda setting. Further, don't see any any evidence of it setting targets and measuring progress or even getting progress reports.

-- Roy Bamford

So, now straight to the point, we could elect a Core Team, including people from each team. And those will be the responsible to take Gentoo into new 'realms', with its 'risks' included. I am also scared about this model .. it might not work, it actually might create the next armageddon for many. But what if it does?, it might help solving this stagnation state Gentoo is facing right now, and bring more new ideas into play.

-- Luis F. Araujo

There's no detail in what you want to do, only a vague unhappiness with how things are, a desire to return to the "good old days" that never were, backed up by arguments that are demonstrably and factually incorrect or incomplete.

What is your plan? Where do you want to take Gentoo, where it isn't already going? ...

_If_ you're looking at Ubuntu with envious eyes, my advice is that you cross the floor and join them. There's no sense whatsoever in putting Gentoo head-to-head with any of the other Linux distros, unless they try to come after what we are good at.

-- Stuart Herbert

As an aside, this has long been the fundamental structural problem in the open source movement. Within a given project, things generally find a way to get done, but when a problem lies between two projects (be they peers, one dependent on the other, whatever) then things often remain unresolved....

This is actually the cutting edge area in the free software movement at the moment - trying to find a common ground for not just projects but constellations of projects and above them distros to collaborate.

-- Andrew Cowie

In this context, it can also be interesting to read Matthew Garrett's note on his departure from the Debian Project:

There's a balance to be struck between organisational freedom and organisational effectiveness. I'm not convinced that Debian has that balance right as far as forming a working community goes. In that respect, Ubuntu's an experiment - does a more rigid structure and a greater willingness to enforce certain social standards result in a more workable community?

The management of large-scale projects is hard - this has been known for centuries (or longer). Free software projects bring in some interesting new factors, however, as a result of their voluntary nature and distribution over a wide range of languages and cultures. We are unlikely to find definitive solutions to issues which have been around so long, but, perhaps, we'll learn some interesting lessons in the attempt.

(Log in to post comments)

Managing Gentoo - a study in quotes

Posted Aug 31, 2006 14:18 UTC (Thu) by landley (subscriber, #6789) [Link]

Replacing an individual with a committee seldom results in faster or
better decisions. A wedge with more than one point does not work.

Open source development works on a publishing model: contributors form a
slush pile and an editor goes through it and says "no" to 90% of it to
fight off Sturgeon's Law. If there's time, the editor polishes up
what's left, stitches it together into a magazine, and puts out the next
issue. On larger projects, you have an Executive Editor (Jonathan
Corbet) who can farm chunks of the slush pile out to junior editors
(Rebecca Sobol, Forrest Cook...) who either handle a module themselves or
pass on the best 10% of what they read to the executive editor, who may
still throw some of what's left or polish it some more before stitching
it together into the next issue/release.

The purpose of the editor isn't to write the articles, the purpose of the
editor is to decide what goes in the magazine. And the editor does this
by saying "no" to anything that's not good enough. The main power at the
top is veto power, but it's still extremely important.

And there's a difference between magazines that take submissions from
anybody, and magazines that have no strong editorial vision.
"Peer-reviewed journals" tend not to be very interesting to read
(although individual articles in them might). A good executive editor
exercises taste and judgement to take calculated risks. This is just
about the opposite of the consensus that comes out of a committee.

I've said for years that Gentoo was replacing Debian. I'm sad to see
they're being a bit more thorough about it than I expected. Who is in
charge?

Managing Gentoo - a study in quotes

Posted Aug 31, 2006 14:34 UTC (Thu) by smitty_one_each (subscriber, #28989) [Link]

>Who is in charge?
Why, the dude on first, clearly. ;)

Managing Gentoo - a study in quotes

Posted Aug 31, 2006 17:43 UTC (Thu) by g2boojum (subscriber, #152) [Link]

> The purpose of the editor isn't to write the articles, the purpose of the
> editor is to decide what goes in the magazine. And the editor does this
> by saying "no" to anything that's not good enough. The main power at the
> top is veto power, but it's still extremely important.

Actually, the existing council (committee) actually does a pretty good job
of vetoing bad ideas. What most of the quotes are suggesting, however,
is that veto power isn't really enough. A number of Gentoo developers
strongly feel that the folks at the top of the heirarchy should be setting distribution-wide goals and guiding the rest of the developers in achieving those goals. In the current system the setting and achievement of goals is mostly limited to individual developers and projects.

> I've said for years that Gentoo was replacing Debian. I'm sad to see
> they're being a bit more thorough about it than I expected. Who is in
> charge?

Ultimately, the people who choose to develop the distribution are in charge. That's always the underlying reality of a community distribution. Volunteers will work on what they find interesting, or useful, or necessary, and they will "vote with their feet" if they are not allowed to work on stuff that fits those categories. Leadership under such conditions is therefore an exercise in herding cats. It may be necessary, but it is by no means simple.

Incidentally, although people often scoff at the notion of leadership by committee, the job of leading a distribution is easily a full-time job for a single person. Since our devs aren't paid (by Gentoo, anyway), that isn't a very realistic strategy.

Managing Gentoo - a study in quotes

Posted Aug 31, 2006 16:07 UTC (Thu) by dberkholz (guest, #23346) [Link]

Here's the current structure: http://www.gentoo.org/proj/en/glep/glep-0039.html

Gentoo's only small if you consider 350+ developers small. It takes a lot of work to maintain a database of more than 10,000 packages, along with all of our documentation, infrastructure, and other miscellaneous projects such as the installer.

Managing Gentoo - a study in quotes

Posted Aug 31, 2006 17:19 UTC (Thu) by g2boojum (subscriber, #152) [Link]

> It would be interesting to have an article on how Gentoo's management
> structure officially works.

The current (official) structure is:
http://www.gentoo.org/proj/en/glep/glep-0039.html

Whether it works or not is (clearly) open for debate.

The previous management structure is documented in:
http://www.gentoo.org/proj/en/glep/glep-0004.xml

Managing Gentoo - a study in quotes

Posted Sep 1, 2006 15:25 UTC (Fri) by djmutex (subscriber, #12657) [Link]

I second (third?) this as well. I've been running Gentoo on my home machine since Jan 2003, and it's the only Linux I haven't managed to break beyond repair within a few months.

With all binary distros I sooner or later ran into some upgrade problem that prevented them from starting and that wasn't trivial to fix, and with Gentoo, if such a problem ever occured, there were the always-helpful forums to my rescue.

Since then, I have used Gentoo on all my computers. That includes my laptop, my office machine, two public web and mail servers (with hardened Gentoo), and my new office machine will have it too.

So all I can say, it is the best operating system I know. For both desktop and server.

To me, if there's one thing where Gentoo clearly needs some steering, it's portage. From having read gentoo-dev for a few months now, this appears to be one of the main issues of contention as well (with the usual personality issues). Even though it's a great system, portage just doesn't scale, and even 2.1 is still unacceptably slow. Gentoo as a whole looks undecided to me about where to go from there.

Managing Gentoo - a study in quotes

Posted Sep 1, 2006 16:16 UTC (Fri) by iabervon (subscriber, #722) [Link]

All of the persistant problems I've had with Gentoo so far have been fixed by leaving it running the way it is for a day or two, syncing, and trying again.

I agree that portage needs work, but I think the main issue is that it needs to be able to simultaneously merge multiple updates (to deal with the case where the state before is fine, and the state after is fine, but there are conflicts preventing any series of individual steps that doesn't include removing a package in between; it only comes up on occasion, but, when it does, it's scary, since you have a period when your system doesn't have /bin/login or something) and it needs to have back pressure in the dependancy solver (e.g., new versions of package A require versions of package B which are masked; treat those new versions of package A as masked, and thus the less-new version of A is the newest acceptable; or new versions of glibc only work with the nptl USE flag, which is not set, so those versions should be treated as masked unless the USE flag gets set).

I do think that it needs the ability to sync only a single package, and it should do this automatically if trying to build it fails (I've had a number of days when version N is current for my middle-of-the-night sync, but N+1 is current and the files for N, which was badly broken, are gone in the morning when I try to install it; it'd be nice if portage would work this out without requiring a manual and full --sync). I find that the speed isn't bad if I sync from a nightly cron job, so it's done all of the slow and uninteractive stuff when I'm not waiting.

Oh, and today's quibble: glibc-2.4 depends not only on having a recent gcc, but on that gcc being the compiler used to build with. Yesterday, it built me gcc 4.1.0, and then aborted trying to build glibc-2.4 with the still-default gcc 3.3.x. Portage ought to know that, if you depend on a particular version of a slotted package, it needs to make sure that you actually use the selected version.