Distributions

Term limits and the Debian Technical Committee

Debian's Technical Committee (often abbreviated as TC or, on Debian lists, ctte) has been in the news quite a bit lately. The TC acts as Debian's final arbitrator in disagreements between project members, and 2014 has seen more than the average number of such disagreements. In addition, some of the debates within the Debian community as a whole have evidently proved to be enough of a strain that several long-serving TC members have resigned from the committee in recent months. Naturally, high-profile technical disputes and resignations from the TC cause attention to turn to make up and processes of the TC itself. On December 1, former Debian Project Leader (DPL) Stefano Zacchiroli proposed a major change to how the TC operates: implementing limited terms for TC members.

An old idea

The idea of TC term limits was raised most recently in May, when Anthony Towns suggested adopting some set of rules that would change TC membership from its current de-facto "for life" appointment to something finite and well-defined. Towns speculated on a variety of possible options without promoting any one option.

Several other project members (including some on the TC) weighed in during the ensuing discussion, and the general consensus seemed to be that there were merits to idea. For one, a never-changing TC could (theoretically) turn into a cabal or simply get trapped in "groupthink" caused by having a limited set of voices. For another, as Russ Allbery noted, the perpetual nature of a TC appointment may be causing appointments to skew toward cautious and conservative choices. In contrast, he said, "I think our DPL selection process works extremely well and benefits greatly from having a yearly election."

But the final major reason for considering time-limited terms is that—as pointed out by Allbery, Towns, and others—the TC's lack of a mechanism for stepping down can make a departure difficult. Towns said "it would be nice if there was a way out of the ctte that had more of a feeling of winning / leaving at the top of the game", while Allbery sought to find a way to give TC members "a clean break point where they can stop without any perceived implications of resigning, so they can either decide they've done enough or they can come back refreshed and with fresh eyes." On Allbery's final point, it is indeed easy to read comments and discussion threads about several of the recent TC resignations and find people speculating on the reasons behind and ramifications of each individual departure.

Nevertheless, the discussion started by Towns about term limits ended without a concrete plan of action. There were concerns about how to implement term limits without making arbitrary decisions about what constitutes "enough" time, as well as concerns about how to implement any term-limiting mechanism without causing undue turmoil—by (for example) immediately losing half of the TC's membership.

A new proposal

Perhaps the turmoil within Debian and in the TC itself over the past few months served to make the prospect of shaking up the TC membership rules seem less intimidating. Or perhaps with several seats opening up on the TC due to resignations, it was simply a good time to consider other changes as well. Either way, in mid-November, Zacchiroli sent out a message proposing a change to section 6 of the Debian Constitution to implement TC term limits. His proposal is a General Resolution (GR), which would require a vote by the entire project.

Zacchiroli's initial draft underwent multiple revisions during the last half of November, but by December 1, he made it a formal proposal. The current version of the proposal aims to set the maximum term for TC members at around four years, but with some flexibility built in to account for resignations and other departures. The goal is to replace two TC members each calendar year, so that all seats on the committee are rotated through every four years. In addition, former members must stay off the TC for at least one year before they can be re-appointed.

The specifics of the wording are worth looking at as well. Each year on January 1, if two senior TC members have served for more than 3.5 years, those two will have their memberships marked for expiration—in other words, their terms will end on the coming December 31. Because new appointments to the TC can happen at any time, there is some variation in how long a "full" term would last; as Towns observed, "the max age is 5.5 years (appointment on Jul 2nd, hitting 4.49 years on Jan 1st, then expiring at 5.49 years next Jan 1st)". Nevertheless, most on the list seemed to find the issues of regular rollover and requiring a one-year "mandatory vacation" (as current DPL Lucas Nussbaum called it) to be the most salient factors: precisely how long anyone sits on the TC is an implementation detail.

Dropped along the way were provisions to prevent the term-expiration mechanism from leaving the TC with less than four members (out of the total of eight seats), various suggestions to change the number of TC seats, and a suggestion that the remaining TC members decide whether or not to re-appoint a member whose term is expiring. Objections to these ideas varied, although the ones that seemed simply too different from Zacchiroli's core proposal (such as changing the size of the TC) were usually dropped on the grounds that proponents should raise them as separate GRs.

Similarly, Clint Adams proposed eliminating the TC altogether. The idea does not seem to have widespread support, although Allbery commented that he had considered making a similar proposal himself in the past—only to decide that whatever dispute-resolution method replaced it would not be any better.

That said, there was considerably more discussion of how the rules could be adjusted to place an upper limit the amount of churn that the TC undergoes each year. This year, for example, three committee members are stepping down; if two additional seats were to expire automatically, then more than half of the TC would be replaced in a single year—an outcome few consider ideal for the health and stability of the project.

Some of the early discussions about the proposal included specifying a transition mechanism to let the current longstanding TC members rotate out gradually rather than all at once. Ultimately, some modifications to the two-senior-seats-automatically-expire plan arose that would throttle the turnover rate, and have the beneficial side effect of making the addition of a transition mechanism into the Constitution unnecessary.

Three alternatives (summarized by Nussbaum) to the original two-seats-expire-per-year plan were proposed. The first, which is known as the 2 − R plan, would have the two seats automatically expire if there are no other departures from the TC, but would subtract from those automatic expirations the number of resignations, retirements, or removals ("R") that happened during the past year—stopping at zero, of course.

The second alternative is a slight adjustment of the first, and is known as the 2 − R′ plan. It would subtract from 2 only the number of resignations or departures of people who would otherwise be candidates for seat expiration (that is, resignations by members with 3.5 years experience or more). In short, this plan would ensure that the resignation of junior TC members would not cause the most senior members to remain on the committee an additional year.

The third alternative, known as 2 − S, is a subtle modification of the 2 − R′ plan. It would subtract from 2 only the number of resignations in the past year by members whose terms would definitely have expired at the end of the year otherwise. That is, under the 2 − S plan, only a resignation by one of the two most senior seats can decrease the number of automatic term expirations. Under the 2 − R′ plan, it would be possible for the third-most-senior member to resign and cause a reduction in the number of automatic seat expirations, if at least three members had been on the TC for longer than 3.5 years.

Such a condition cannot arise when there have been several years of two-seat rotations in a row, of course. But it happens to be the case now, since so many of the existing members have been on the committee for a considerable length of time. And more importantly, as Raphaël Hertzog pointed out, it can happen again if there are several resignations (followed by several appointments) in the same year.

If one happens to find the distinctions between the various expiration formulae less than perfectly clear, fear not. Nussbaum outlined the practical effects of the main plans (the original, 2-seat plan and 2 − R). Under the original plan, Bdale Garbee and Steve Langasek's terms would expire on January 1, 2015. Subsequently:

While under 2 − R, the resignations already announced in 2014 would mean no additional seats expire in January 2015, after which:

The differences in the long term are, to be sure, subtle enough that most assessments of which plan is better will boil down to personal preference. Ultimately, Nussbaum added the 2 − R option as an amendment to Zacchiroli's proposal.

What's next

Zacchiroli's proposal quickly garnered enough seconds to move it forward for a vote. As per project procedure, at least two weeks of discussion will follow, after which any of the proposal's sponsors may call for a vote.

There seems to be little resistance to the idea of rotating TC members more frequently—if nothing else, to prevent burnout among qualified project members. But the term-limit idea would constitute a major change in how Debian functions, which is a notion that makes many people uneasy to one degree or another.

On the other hand, the main objection to too much rotation within the TC is the hard-to-define notion that it would weaken the project. Towns, for his part, contended that the idea of "newbies" on the TC causing weakness to Debian are "at the far end of hypothetical". There is, the argument goes, not a shortage of project members who would make positive contributions to the TC, and new committee members will still be selected by the sitting TC with the approval of the DPL. So fears about a TC composed of unqualified people apt to make poor, reckless decisions are unfounded.

The discussion process is taking place on the debian-vote mailing list. Whenever the final vote itself takes place, the outcome will be announced there as well. Although the exact form of the process has yet to be decided, the way things stand today it seems likely that Debian will soon have a formal process in place to regularly rotate members in and out of its top decision-making body.

Comments (6 posted)

Distribution quotes of the week

Comments (57 posted)

The "Devuan" Debian fork

Comments (440 posted)

BSP in Switzerland (St-Cergue)

Full Story (comments: none)

Fedora Council election results

Full Story (comments: none)

Fedora 21 betas for ARM and POWER

Comments (none posted)

Announcing openSUSE board election 2014/2015

Full Story (comments: none)

Distribution newsletters

• Debian Misc Developer News (#37) (November 25)
• Debian Project News (December 1)
• DistroWatch Weekly, Issue 586 (November 24)
• DistroWatch Weekly, Issue 587 (December 1)
• 5 things in Fedora this week (November 19)
• 5 things in Fedora this week (December 2)
• Ubuntu Weekly Newsletter, Issue 393 (November 23)
• Ubuntu Weekly Newsletter, Issue 394 (November 30)

Comments (none posted)

They make Mageia: David Walser (Mageia Blog)

Comments (none posted)