The story of Gentoo management

Early managers

The first known well-defined form of Gentoo management was the top-level management structure (apparently later called metastructure) defined in GLEP 4. The structure is described as:

[...] This management structure will consist of the chief architect and a group of developers that will be given the title of "Top-level managers." Top-level managers will be accountable for the projects they manage, and be responsible for communicating the status of their projects to the rest of the top-level managers and chief architect, among other things detailed later in this document.

All the top-level projects in the Gentoo project will be clearly defined, including with goals, sub-projects, members, roadmap and schedules. [...] 2

So apparently a number of top-level projects is defined, and managers (leads) of those projects form the management team along with the chief architect of Daniel Robbins. GLEP 4 also includes the list of initial managers, top-level projects and a lot of bureaucracy for them to follow.

According to the dates inside the GLEP, it was posted by the end of June 2003, and the implementation work started soon after. The first project list was posted on the site on 2003-07-19 3; it has been later renamed (without changes) and can still be found in the archival version of our website 4. It introduces a few differences from the GLEP: gentoo-linux project is gone, and three new projects appear: metastructure, gentoo-base and documentation. It also features a group of orphan projects that apparently are not included in the top-level management.

The original projects page does not clearly distinguish managers from regular project members. In some cases, the top-level projects list only one member, or the first member is listed out of order (with other members being sorted lexically). By comparing the list obtained from those apparent leads with GLEP 4 on one hand, and with future manager lists on the other hand, we can arrange a working hypothesis on who managed most of the projects. According to this hypothesis, the main difference between GLEP and the site would be reducing most of the projects to one manager, with Daniel Robbins disappearing from the explicit listings.

The next data point is 2003-09-22 when the previous static listing was replaced by an index dynamically including different project pages in Gentoo (that was used until 2013) 5. Initially, this file included an explicit list of managers and only three projects (including a new Desktop TLP). According to commit messages, all missing TLPs were added on 2003-09-30 6, forming a list equal to the one from previous site, plus the new Desktop project. A commit later, on 2003-10-09 the last manager was added to the list 7.

At this point, I think that we can form a first clear list of top-level managers in Gentoo. Some of the projects already have a project page with complete member list. Others are only included on the list along with their leads. By combining the managers and leads listed in the top-level projects, we can account for all managers explicitly listed on the project list. However, we do not have enough data to clearly assign them to specific projects yet.

How were managers selected?

One of the key issues on which my research was focused was determining how were various management bodies selected. After all, if managers are to be considered Gentoo's body of self-government, it would be crucial for Gentoo developers to have an actual say in selecting the managers.

The metastructure proposal was written by Daniel Robbins, the chief architect of Gentoo. While this would effectively put him as the one choosing initial management team, there is evidence to suggest that he based the list on existing responsibilities within Gentoo. Quoting his reply to feedback on the metastructure proposal:

[...] Meaning that existing roles and responsibilities of developers within the Gentoo Linux project will be fully respected and reflected in the organization as this plan is implemented. 8

For example, you can see that the manager of the Hardened project used to be its lead prior to GLEP 4 10. As a matter of historical curiosity, the Hardened project seems to be the first Gentoo project to have a site in the GuideXML system.

The procedure for manager changes afterwards is less clear. It was not specified in GLEP 4. A mail from Daniel Robbins, posted on 2003-07-15 states:

Currently, new managers are elected by unanimous vote of the existing managers. [...] 12

When the Desktop TLP project was formed, the managers apparently had a problem selecting the lead. The problem was pointed out during the 2003-12-15 manager meeting:

Paul de Vrieze (pauldv) presented the nominees for the desktop lead position (foser, liquidx, spider, spyderous, and tseng), and it immediately became clear that the new procedures for electing a manager were far from well-defined.

[...] Dennis M.D. Ljungmark (Spider) from the desktop group pointed out that the desktop group had not really participated in the selection process, noting that "the whole process seem[ed] opaque and closed", a conclusion that apparently rendered, for them, the nomination process itself fairly pointless.

That revelation pretty much shut down any actual discussion of the vote itself, as the discussion then focused on how to construct a fair, serviceable top-level manager selection procedure. [...] 11

The 2004-05-17 Gentoo Weekly Newsletter indicates a change in the selection method:

Another Gentoo Managers' Meeting was held today on May 17th. The first items on the agenda were votes on requiring a supermajority of managers (66%) to confirm new managers, and on confirming John Davis as the lead for Release Engineering. The supermajority requirement was ratified, and subsequently John was confirmed as the the Release Engineering lead. 13

The model of GLEP 4 is hard to judge. On one hand, it certainly ratified delegating some of the decisional power to the developers. On the other hand, the manager selection was always limited to the existing managers and the manager terms were unlimited. As a result, distribution of power still depended on the few people holding it. John Davis has pointed out those problems in his mail dated 2003-07-15 14. It does not seem that his calls were answered at the time.

The rise and fall of Gentoo Managers

The early period of Gentoo Managers is marked by high activity. GLEP 4 required status reports twice a week and weekly IRC meetings. Sadly, a lot of the related data is lost to Gentoo developers since it took place on gentoo-managers mailing list. While we seem to possess the complete archives of it, we can't publish them since there is no indication that the mailing list was intended to be published as a whole. Quoting another of Daniel Robbins' replies from 2003-06-25:

The gentoo-managers list is intended for "meetings." I fully support having the weekly manager status updates posted publicly on project pages as part of our accountability to our users. I think the meetings themselves should be private though. But each project's and subproject's weekly status should be public information.

On the plus side, it seems that the manager meetings were public after all and the logs of all meetings between 2003-11-03 and 2004-07-19 are available 15. Additionally, meetings up to 2004-03-08 include (GuideXML) summaries. Besides that, some of the decisions are to be found in the Gentoo Weekly Newsletter 16.

Did Managers have real deciding power? It appears that they did. Quoting Paul de Vrieze's mail from 2003-11-19:

Key decisions are made by the management team. A formal voting system is being developed, but currently we work with consensus vote. Note that this is consensus of the management team. In this Daniel [Robbins] has some extra edge as the project leader although it is not formal. 17

The first major scratch occurred around 2004-04-26. This day, Daniel Robbins announced his resignation from development roles 18. No Chief Architect was appointed in replacement. GLEP 4 was still in force, although the discipline seemed to fade away. Managers list traffic decreased, Manager meeting logs stopped being published. Once Gentoo Foundation was formed, many of the managers started doubling as Trustees, making it harder to exactly determine the role of Managers in the following year.

The first end point on the timeline of Managers is 2005-06-14. This day, the results of vote for the new metastructure were announced 19. What was later codified as GLEP 39 has been approved 20. Managers were at their last.

The first Council election results were published on 2005-09-01 21. The first Council met at 2005-09-15 22. A few days later, the list of managers was removed from the site 23.

Putting it to the table

Fitting Managers into the management table was not an easy task. But why do it in the first place? I wanted to include them because they were predecessors to the Council. I think the evidence to that is clear — their powers were ratified by GLEP 4 metastructure, and the Council was formed through a metastructure proposal replacing it. In this context, it seemed really interesting to compare the first elected Council with the final team of Managers.

But who were on the final team? This is really hard to answer. When originally researching managers, I had my share of doubt whether GLEP 4 was actually followed to the letter and new TLP leads were becoming managers. While we have found evidence to support that, it is unclear whether it was still taking place during the last months of Manager activity — whether the new TLP leads were confirmed by the Managers and accordingly added to the mailing list.

Let's look at the CVS logs of various project pages in reverse chronological order. We see that Mike Frysinger became co-lead of base system in August 2005; a month earlier Brandon Hale was removed from desktop (apparently inactive since March 24) and Sven Wegener created a new page for QA project listing himself as the lead. All of this happened already past the new metastructure vote, and past last activity of gentoo-managers list. Should we consider either of them Managers then?

Let's look further. May 2005 brings Jon Portnoy (temporarily) stepping down from Developer Relations top-level management and a complete leadership change in Portage project — three new leads. Apparently they weren't added to the -managers list but posted the project status by proxy. I think we can consider this the most relevant point for the final team — although it is unclear if all TLP leads were actually participating in the management, or considering themselves Managers.

Once we know whom to put, the next problem is how to put the data in the table. Originally I wanted to place Managers just below the Council, to emphasize on the succession. However, this met two problems. Firstly, it would put Managers in 2004/05 term (respective to the Trustee term) which would be quite imprecise given that Managers were not really running in terms. Secondly, the count of 20 apparent Managers in May 2005 would be a hard fit, compared to 7 Council members. Therefore, I've decided to place them in a separate table.

Finally, I was wondering whether I should assign them to their respective projects instead of using a flat list. Both approaches have their advantages. A flat list puts more focus on the management team and its members. The project assignment indicates which projects were considered top-level, at the cost of duplicating some of the managers. For those reasons, I've included both variants.

Conclusion

The GLEP 4 metastructure was probably the first official management structure of Gentoo. It structured the distribution in a hierarchy, with Chief Architect and top-level project managers on top, and subprojects below them. It created a certain vertical model, with subprojects answering to their parent projects, and parent projects representing their subprojects.

Was it a revolution? I wouldn't consider it so. It is apparent that some developers already had some degree of decisional power and influence in Gentoo. Daniel Robbins admitted that the structure was meant to reflect the developer roles at the time 8. However, the choice of top-level projects may have reduced the influence of some of the developers. For example, in reply to Joshua Brindle Daniel Robbins admits that his project has been moved under gentoo-alt umbrella 9.

Was it a democracy? Certainly it was a step towards it, though it wasn't one yet. The power started being distributed but the access to it was limited to a closed group of Managers. A few developers have managed to join the group but it certainly wasn't open. In the end, only people contributing to specific areas (top-level projects) could have become Managers, upon approval from other Managers.

Was it a good model? Hard to tell. Certainly it worked for some time but it seems that it eventually started to decline. That might have been related to Daniel Robbins leaving, or to Managers focusing on their other roles as Trustees. That might simply have been caused by people having less time to work on Gentoo. In any case, it eventually stopped fitting the needs of Gentoo at the time and was replaced.

The beginnings of the Foundation

Originally, Gentoo has been backed by a commercial company Gentoo Technologies, owned and run by Daniel Robbins. Apparently, not all developers agreed with this model.

The earliest mention of a Gentoo not-for-profit I was able to find was in Daniel Robbins' mail dated 2003-06-25, in reply to the GLEP 4 metastructure proposal:

On Tue, Jun 24, 2003 at 09:15:00PM -0500, Joshua Brindle wrote:

[...]

2) I didn't see not-for-profit mentioned, don't you think this is essential to the success of gentoo?

Yes, I think it is. Having me move out of the day-to-day management efforts will allow me to focus efforts on getting the not-for-profit started. Right now I am simply too overwhelmed with work. 9

The mails around the time indicate that the non-profit status was discussed and agreed on for some time already. However, it did not seem to be moving forward as some of the developers desired. The for-profit endeavors made by Daniel Robbins have apparently caused it to be forked into Zynot 25.

The work on actually forming the non-for-profit seems to have started in April 2004. The first complete proposal is dated 2004-04-16:

Gentoo Foundation, Inc. proposal

The purpose of this foundation is to hold the intellectual property of the Gentoo free software project. It will have a board of trustees. This not-for-profit will be an open membership trade association.

Trade associations (unlike charities) can be more restrictive in their requirements for membership. Membership will be limited to Gentoo developers. The criteria for being a Gentoo developer will be determined by the board of trustees. There will be no membership dues.

There will be an initial board of trustees appointed, which will be selected to meet my commitments to existing managers and developers. This initial board of trustees will serve for one year from the establishment of the Gentoo Foundation, after which point the board will be elected by the members (Gentoo developers.) After that, regular elections will be held (election cycle TBD) to determine board members.

The Gentoo Store will pay for the establishment of this not-for-profit. The Gentoo Store will also pay for the Gentoo Foundation's application for 501(c)(6) federal trade association status (~$5000 or so?)

Gentoo Technologies, Inc. will transfer the copyrights and trademarks to the Gentoo Foundation. In exchange, the Gentoo Foundation will grant Daniel Robbins & Gentoo Technologies, Inc. perpetual, non-exclusive, royalty-free use of the "Gentoo" trademark and "G" logo. This will allow me to continue to run the Gentoo Store if I want.

I will be a member of the initial board of trustees, to give legitimacy to the Gentoo Foundation and also show my commitment to the future success of this entity. 26

Later on, Daniel Robbins resigned from this role on the initial Board. The Foundation has been established on 2004-05-28. This date, along with the original appointed Board of Trustees can be found in the Articles of Incorporation 27. At this point, the 13 Trustees started working on solving various legal issues regarding the new NFP.

The initial issues found in Sven Vermeulen's 2004-10-21 status update include: registering Gentoo trademark and logo, setting up a bank account, resolving copyright assignment issues and handling donations 28. By December, the work on Bylaws has apparently started 29.

The first quarter of 2005 does not leave any trace of activity on the Foundation mailing lists. However, the first draft of Bylaws is published 30. The first topic of April are upcoming Trustee elections:

The Gentoo Foundation, Inc was incorporated on 28 May 2004. According to the articles of incorporation, a new board of trustees needs to be elected within 13 months of its incorporation, which is coming up fast. At the moment the only members of the foundation are the current trustees, and clearly that needs to change before we hold elections. 31

The Trustees agreed on inviting all Gentoo developers who have been developers for at least one year 32, discussed the election procedures 33 and give birth to the voting software still used today 34.

On 2005-05-21, Grant Goodyear confirmed receiving copyright and trademark transfers from Gentoo Technologies 35. On 2005-09-14, he confirmed that Gentoo trademark has been registered to the Foundation 36. The end of 2005 concludes the early active period of the Foundation.

The dark years

The year 2006 brings little visible progress in the Foundation affairs. This year's Trustee elections bring fewer nominees than the previous ones. Quoting Grant Goodyear's reply to Diego Pettenò:

Diego 'Flameeyes' Pettenò wrote: [Tue Jul 04 2006, 01:34:02PM CDT]

What happens if we don't get any nomination? I know it's early but... well usually there are some early shots, and in 4 days we had nothing. This is strange to say the least.

Well, it will make thinning out the list of trustees much simpler.

[...]

Assuming nobody complains too much, let's assume that the nominations and elections for trustees will coincide with those for council members. I'm still suggesting that we limit the number of trustees to 5, assuming that we will quickly replace any trustees who depart during his or her term. As for Foundation membership, that's still a tricky issue, since the bylaws were never approved [...]. 37

The nominations bring only 5 nominees. Therefore, they are accepted as the new Board without a vote. Seemant Kulleen announces it on 2006-09-05 38. His mail immediately meets strong response from Gentoo developers who disagree with skipping the vote.

On 2006-09-23 a new election is announced 39. This time, 7 developers are nominated and five of them are elected into the Board (with small changes, compared to the earlier result) 40. However, Seemant Kulleen resigns almost immediately 41, and Stuart Herbert a month later 42.

There is no apparent Trustee activity during the first half of 2007. The single mails sent by Gentoo developers to the mailing list remain unanswered. The upcoming 2007/2008 election does not bring even sufficient number of candidates to fill all the slots 43. At this point, Trustees seem to be working towards SFC inception 44. There are no further news until 2008.

The first massive peak of activity results from Daniel Robbins' 2008-01-11 blog post stating:

Several days ago, the Gentoo community discovered the unfortunate news that the Gentoo Foundation's charter has been revoked for several weeks, which means that as of this moment the Gentoo Foundation no longer exists.

[...]

It also appears that all but two of the interim Foundation trustees have either resigned or are unreachable. Grant Goodyear appears to be the only remaining trustee who actively does legal stuff, along with Chris Gianelloni who runs the Gentoo Store. 45

The Grant Goodyear's update from 2008-01-18 provides more details:

With help from Renat Lumpau (rl03), I spent some time this week talking to the Foundation's lawyers, collecting documents, and sifting through old e-mails. As I posted on gentoo-nfp a couple of days ago, the state of New Mexico did, indeed, revoke the charter for the Gentoo Foundation, Inc. in October of 2007. It's still not entirely clear why, since I mailed a check along with the (then) current and past-due annual reports to the state of NM way back in July. Since the check never cleared, it seems a good guess that the paperwork went astray, but we won't know until Renat's request (and $5) are processed by NM and they get back to him. 46

He also confirms that all Trustees but himself and Paul de Vrieze have resigned.

By the end of the month, nominations for new Trustee Board start. This time, a full Board of 5 Trustees is elected from the initial 8 nominees 47. The new Trustees are much more successful.

By 2008-05-13, the Foundation is reinstated 48. On 2008-08-31, the first Bylaws are adopted 49 50.

Foundation then, Foundation yesterday

Since then, the Foundation has been operating with no major interruptions. The initial Bylaws already included the Trustee election model still used today:

Trustees shall normally hold office for a period not exceeding two electoral periods. Trustees shall retire annually by rotation (and may be re-elected). [...] 49

They also included the restriction that:

No individual shall serve as a Gentoo Foundation Trustee and Gentoo Council Member concurrently. 49

Initially, only Gentoo developers were admitted to the Foundation. However, the 2008-10 meeting already changed them to allow non-developer members 51:

Active Gentoo developers who are not members of the Foundation may apply for membership. Any developer applying for membership in the Foundation will become a member of the Foundation immediately after the next Trustee meeting following the application unless an absolute majority of the trustees (currently 3 out of 5) oppose membership for the developer at this meeting.

Applicants who are not Gentoo developers need to cite verifiable evidence of contributing to Gentoo or to the stated aims of the Gentoo Foundation Inc. 52

The two following years again form a period of low activity, built mostly around Trustee elections. The silence is broken in January 2011 by a series of posts by William L. Thomson Jr., apparently summarized in his final mail:

Just making a quick list of the items/questions asked buried in a few posts in a previous thread. No need to reply till there are answers. This is just a list/summary of the items/questions to be answered once the internal audit has been completed.

I am very appreciative and thankful that the trustees will be conducting an internal audit. I will patiently away the outcome, and answers to the following items/questions.

1. State filings, with NMPRC and State Attorney General, bi-annual filings with NMPRC, and annual with State Attorney General

2. Federal filings, annual tax return, form 990 (990-N or 990-EZ)

3. Where is the money from the old bank account? When was it received/deposited into new account, and how much?

4. What is the cause behind the $9k discrepancy in Q3 2010, and why was that not caught sooner, like in 2008 or 2009? 53

In November 2011, Matthew Summers indicates that Foundation filings are in order.

By the end of 2012, the topic of copyright is brought again, by Rich Freeman 55. This is not the first time Trustees debate about copyright. However, I think it is worth mentioning since today's copyright work is based on the draft policy written by Rich.

Over the following years, the NFP mailing list traffic is low. Recurring topics include copyright, logo usage and Social Contract, financial reports and elections.

The recent months

The Foundation activity starts to peak again in October 2016. Matthew Thode opens a discussion on changing the procedure for admission of members to include a quiz 56 which does not eventually get implemented. Neither does Robin H. Johnson's proposal of admitting developers automatically 57.

On 2016-11-07, Alec Warner proposes merging Gentoo developers and staff members into a single Gentoo member type 58. This is probably the first case in years when Trustees reach outside the direct affairs of the Foundation. In reply, Michael Palimaka points out that:

That is no longer correct - "staffer" is a thing of the past. These days, everyone is a developer whether they work on ebuilds or not.

[...]

While I applaud your efforts, the proposal seems to be based on an outdated picture of the community. Additionally, given our current metastructure, it's not clear to me how this is even a Foundation issue. 59

On 2017-01-05, Matthew Thode proposes:

[...]

In order to solve this Gentoo needs to have a combined electorate, meaning those that would vote for Council would also vote for Trustees and visa-versa. This would ensure that everyone’s needs are represented. We should have a single combined governing body, let’s call it ‘The Board’. This is so that conflicts between Council and Trustees (as they exist now) would have a straightforward resolution. This new ‘Board’ would be able to use the existing project metastructure to delegate roles to various groups (Comrel, Infra, etc would still exist, but under this new Board). 60

This post has started a major discussion on the role of Trustees and Council that is still running today, as well of the purpose of the Foundation. It brought many replies, counter-proposals and reiterations. Since the process is far from finished, I will not be getting into details.

The last fact to mention is that on 2017-06-19 Bylaws have been changed again, to add the following requirement for Trustee candidates:

Candidates standing for election must be active Gentoo Developers as of the record date (Effective 2017/07/04). 61

Trustee elections

While the first Board of Trustees has been explicitly appointed by Daniel Robbins, the subsequent Boards are subject to elections. The elections occur annually, and consist of nomination phase followed by voting. The votes are counted using Schulze method (which is one of the Condorcet methods).

The electorate consists of all Foundation members. Originally, the membership was offered to all project members who were Gentoo developers for at least one year. By October 2008, this was changed to allow all Gentoo developers (without the one year limit) and non-developers who have contributed to Gentoo. According to pre-2008 Bylaws and traditionally past that, Foundation membership is terminated after not voting in two consecutive Trustee elections.

At first, all Foundation members could have been nominated. The 2008 Bylaws change therefore implied opening Trustee nominations to non-developer Foundation members. This was changed again in 2017, restricting Trustee candidates to active Gentoo developers. Since 2008, the Bylaws have explicitly forbidden a single individual from being a Trustee and a Council member simultaneously.

The first election, held in 2005 resulted in 13 Trustees being elected. 7 of them were reelected from the original Board. Upon the failure of 2006 nominations, the Board was reduced to 5 Trustees, and their term was ratified to the period of two elections. Starting next year, half of the Board was supposed to be rotated annually. However, in reality this did not start happening until 2009. Afterwards, in a few cases Trustees served for three years without reelection.

The popularity of Trustee elections fluctuated over the years. As you can see in the table, only some of the elections actually involved voting. Three of them (if we skip the one which was restarted) were limited to the nominations phase, as they did not bring enough nominees to dispute any of the seats. The 2007 election brought so few candidates that the Board did not change at all.

Besides annual elections, Trustee seats were subject to mid-term resignations. Those seats were filled by vote among the remaining Trustees. There were three kinds of candidates selected this way:

• highest-ranked candidates which did not make it to the Board during the previous election (H in the table),

• Foundation officers at the time (O in the table),

• Gentoo developers replying to open recruitment notices (R in the table).

Nevertheless, the Trustee Board historically had empty seats during the 2006-2008 period, and the second half of the 2008/09 term.

The Trustee elections certainly had a democratic bit to them. However, they have a few weaknesses. Those are:

1. Lacking a mechanism to veto a nominee, especially in case there are too few to hold an election.

2. Explicitly allowing arbitrary Trustee selection in case of vacant seat.

3. Giving Trustees arbitrary, direct control over their electorate.

The first argument was already raised during the first 2006 election where all nominees were to be appointed Trustees without a vote due to lack of candidates. Ciaran McCreesh already tipped a solution back then:

A Debian-style "reopen nominations" option with a vote would make more sense... 68

However, this proposal has never been considered seriously until I proposed a more complete scheme that combined it with fallbacks that accounted for the apparent necessities of the existing system and provided an accurate timeline 69. This proposal has been accepted during May 2018 Trustee meeting and hopefully will be used during the election following it.

The second point can't be solved that easily. As some of the Trustees point out, it is important to avoid long-term vacancies in the Board. Therefore, Trustees will continue to appoint replacement for resigning Trustees and hold power to appoint new Trustees in case of vacant seats. On the plus side, those nominations are usually based on the previous elections, so they end up being at least partially democratic.

The third point is specific to the simple structure of the Foundation. At the moment, Trustees both approve (or reject) member applications, and (via post-2008 Bylaws) determine the criteria for loss of interest which influence membership terminations.

Finding the Trustees

The previous section was focused on the theoretical side and examples of Trustee elections. Now it's time to say how I've managed to obtain the results that can be seen in the table, what data sources did I use and what that choice implies.

My first idea was to use the results provided by the Elections project 62. However, this included only the results for the three most recent elections, and full Board member lists only for two of them. By looking at the git repository 63 or the historical CVS repository 64, I could obtain some more data but it would still not be sufficient to form a full view.

My second idea was to use the Foundation website history. However, the effort of checking both the Wiki page history 65 and the historical GuideXML page history 66 only revealed more issues than it solved. Not only was the data limited to changes in the Board with little to no explanation but also some of the delays and partial updates made the data confusing (e.g. Roy Bamford has been listed as a Board member for 3 months after new Trustee was added).

Nevertheless, the website history helped me a great deal as a secondary information source which I consulted to confirm my results and establish the approximate dates needed to find evidence of Board changes.

The third and final source of information were the Foundation mailing lists. Roy Bamford helped me greatly by suggesting to look into the election announcement mails that indicated which Trustees were being replaced by rotation. At this point, the painstaking work started. With each election representing a time point, the election results mail established part of the Trustees past this point, while the election announcement established part of the Trustees before this point who finished their term.

This approach did not work reliably for all the terms though. Mid-term Trustee changes and Trustees serving three terms made the data inconsistent, and confused me thoroughly. However, with the help of recollections of past Trustees combined with thorough search over mailing list archives and comparison with website changes, I was eventually able to establish the historical Board members and their approximate terms.

Two periods posed the biggest problems to me: the 2006-2008 period, and the 2014-2016 period. In the first case, the limited data confused me deeply. I was able to find the 2006/07 election announcement, the two resignations and the replacements. However, I wasn't able to find anything complete on the 2007/08 election. Only after being pointed to Grant Goodyear's 2008 status mail 46, I was able to establish the state of affairs by the end of 2007/08 term. The period in between is still marked as ‘missing data’.

In the case of 2014-2016, the data I've collected simply did not fit. I've requested help on the Foundation mailing list 67. It turned out that my problem consisted of two successive cases of three-year term combined with mid-term resignation.

Once I had a complete table of Trustees, the remaining task was to establish how they were elected. Based on the election mails, I was able to distinguish full elections from nomination-only elections. For mid-term replacements, the combination of meeting logs, election mails, other mails and site logs led me to determine the most likely reasons why a particular individual was selected.

Lastly, I would like to list one more source of information: meeting logs. While working on the Council member lists, I was able to use their meeting logs to successfully verify the Council members throughout the term. Sadly, the Trustee logs were far from helpful. Neither the summaries (which rarely contain any useful information, and look more like agenda), nor their metadata contained the list of attendees. Even worse, the roll calls during the meetings were hardly readable and did not really distinguish current Trustees from other attendees who decided to ‘wave’ their hands along with them.

Conclusion

The Gentoo Foundation has been formed in 2004 to provide a not-for-profit entity to hold Gentoo copyright and trademarks, process financials and handle all other legal aspects of Gentoo's existence. Since then, it had its ups and downs.

They had to face a number of problems, starting with the periodic lack of interest in Trustee positions and limited interest of third parties in helping Gentoo. They have reached to umbrella organizations more than once but without success 44.

They had to handle accounting, and sometimes deal with the problems left over by their predecessors 53. They had to deal with legal and IRS filings. The former has already caused them trouble 46, the latter is still not completely solved as of today 70.

Historically, Trustees were handling issues directly related to the legal and financial support of Gentoo. This included such topics as copyright issues (which are handled by a dedicated team today), trademarks, donations and spendings. However, recently Trustees started showing interest in gaining influence over wider aspects of Gentoo, with proposals going as far as to change the metastructure 60.

The Trustees have once again reached the peak on the timeline of their activity. Gentoo is at an interesting point, and it is hard to predict what the future might bring. Hopefully, all the most important problems will be solved during this year, though unlikely before the next Trustee elections.

The metastructure debate

The history of Gentoo Council begins in 2005. It being with a poll on changing the Gentoo metastructure as posted on 2005-06-08 71. Sadly, it seems that all the discussion leading up to the vote was kept private. However, we have been able to extract the proposals put up for vote and Ulrich Müller has prepared a MediaWiki conversion of their texts. The proposals put up for vote were:

• The “FOSDEM” proposal, splitting projects into 7 groups, requiring quarterly project reports and manager meetings with one representative from each project 72.

• Thierry Carrez' “Alternative” proposal, also splitting projects into groups, with project-level meetings, group-level meetings of project leads, top-level meetings of project group secretaries and obligatory GLEPs for every global change 73.

• Grant Goodyear's “Oldschool” proposal with non-grouped projects and a small Council (7-13 members) 74.

• Grant Goodyear's “Oldschool” proposal with a large Council (e.g. ~10% of developers) 74.

• Ciaran McCreesh's proposals combining both “Oldschool” proposals with “boot for being a slacker” 74 75.

• Keeping the GLEP 4 metastructure 2.

• Jason Stubb's “task force” proposal, combining the GLEP 4 metastructure with a non-voting, top-level Task Force to “document and improve all structure, policies and procedures.” 76

The vote results have been announced on 2005-06-14. The “Oldschool” proposal with “boot for being a slacker” and a small Council won 19. The final version of the proposal has been codified into GLEP 39. It describes the Council the following way:

[...]

2. Global issues will be decided by an elected Gentoo council.

   • There will be a set number of council members. (For the first election that number was set to 7 by acclamation.)

   • Council members will be chosen by a general election of all devs once per year.

   • The council must hold an open meeting at least once per month.

   • Council decisions are by majority vote of those who show up (or their proxies).

   • [...]

   • Disciplinary actions may be appealed to the council.

20

The first Council meeting was held on 2005-09-15 22.

Council elections

The nominations for the first Council do not seem to have been public. Apparently, the voting was open throughout August 2005, and the results were published on 2005-09-01 21. However, in 2006 the nominations were already public 77.

The same model as for Trustee elections is used. The elections occur anually, and the Council term lasts a year. The elections consist of a nomination period, followed by vote. Votes are counted using Schulze method.

The electorate includes all active Gentoo developers, and only Gentoo developers can serve on the Council. New Gentoo developers are admitted through Recruiters project. The developer status can be terminated by Undertakers project because of inactivity, or by Community Relations project as a disciplinary action (historically, both those functions were handled by Developer Relations).

When a Council seat was vacated throughout the term, the Council members either have voted on accepting the next highest-ranked candidate from the previous election, or ran an election for the missing seat. GLEP 39 explicitly specifies procedures for removing inactive Council members and reelecting the Council when a meeting does not reach the quorum of 50% 20. Of these procedures, to this day the former has been used exactly once (2009-12).

The _reopen_nominations option has been introduced during the second elections of 2008 (for the missing seat):

An important point in this election is the new _reopen_nominations candidate. If this candidate ranks over all other candidates, the election will be reopened. Any candidate that ranks below this candidate won't be taken into account if there's a need to replace any member of the council until the end of term. 78

This finalized the Council election model as it is still used today. Let's look at the same aspects that I've listed as problems in Trustee elections:

1. Council elections do provide a way to veto a candidate. If his name is voted below _reopen_nominations, he will not take the seat.

2. The Council members have been appointing members for mid-term vacant seats. Only in two of the cases an election has been held.

3. The Council does not directly manage its electorate. This is done by separate privileged team. However, the Council does have indirect influence via serving as the appeal court for disciplinary actions. There is also no limitation in Council members doubling as members of those teams.

The Council elections seem to be more democratic than Trustee elections then. The problem of vetoing a candidate has been solved there in 2008 already, and the wider structure reduced the Council's control of electorate.

Important points in Council history

The history of Gentoo Council is not as interesting as the history of Trustees. During all the past years, the Council has been deciding on technical and management aspects of Gentoo, reviewing GLEPs, approving EAPIs, serving as a final court for technical disputes and an appeal court for disciplinary actions.

Andreas K. Hüttel has compiled the past Council meeting summaries into a single PDF document 79. Nevertheless, I'm going to note a few events that had affected Gentoo long-term.

2006-01-12

Council members are forbidden to act as proxies for other Council members 80. GLEP 39 is updated appropriately 81.

2007-03-15

The first version of Gentoo Code of Conduct is approved 82.

2007-06-14

gentoo-project mailing list is introduced, originally as unmoderated alternative to gentoo-dev 83.

2007-10-11

GLEP 39 is amended to require a request-for-comments from every new project 84 85.

2008-09-11

PMS/EAPI 0 is approved 86.

2010-08-09

gentoo-council mailing list is disbanded, in favor of gentoo-project 87.

2014-01-19

GLEP 39 is amended to account for project pages being migrated to Wiki 88.

Establishing Council members

I have to say that of all management bodies considered, establishing the Council members was the easiest task so far. Thanks to single-year terms combined with yearly elections, I was able to assemble most of the table based on the published Council election results 89 90.

The Council meeting logs page includes summaries of all Council meetings along with list of participants (Council members at the time) 91. Using this list, I have verified my earlier results and established all the remaining mid-term seat changes, the latter confirmed by actual meeting logs.

That's all. I have easily assembled the complete list, and verified each row with two sources.

Conclusion

The Gentoo Council has been formed in 2005 as a part of new metastructure. It represents a shift from a hierarchical structure where global decisions were made by privileged top-level project leads to a system where Council is formed independently of project management, and creating new projects is trivial.

The Council has dealt with many global issues, both technical and social in nature. It has been reviewing and approving various proposals, including GLEPs, EAPIs (final approval) and other community requests. It has been asked multiple times to resolve technical conflicts, override project decisions and review disciplinary actions. However, it has never interfered with the Foundation affairs.

The majority of issues brought to the Council were technical in nature, and that may have brought some developers to believe that the Council is limited to deciding on technical issues. In many cases, the requests for the Council (that were not appeals) overlapped with the scope of other projects in Gentoo, most notably Quality Assurance team.

The -nfp proposals of 2017 60 have started a new discussion about the role of the Council in Gentoo. It is possible that the developers will eventually have to revisit the Gentoo metastructure and reconsider the relation between Gentoo Foundation and Gentoo developers.