The incubator is currently of a scale that means it can no longer operate as a standard consensus driven PMC. It is not that much smaller than the TLPs part of the foundation. Perhaps it would make sense to see how the model that has scaled well for the foundation can be applied here:
ASF Members elect a board board delegates to PMCs (no micromanagement, just delegation of broad goals) PMCs answer to the board Board answers to the Membership PMCs delegate to committers (no micromanagement just broad goals) Committers do the work and make localised decisions Committers are answerable to the PMCs Why can't the IPMC work like that? Well, to a large extent it does. Here are the same items expressed from the perspective of the IPMC and its relationship with PPMCs. No equivalent of "ASF Members elect a board" IPMC members delegate to the PPMCs (including Mentors) PPMCs answer to the IPMC IPMC answers to the Board PPMCs delegate to committers (no micromanagement just broad goals) Committers do the work and make localised decisions Committers are answerable to the PMCs Where this model breaks down is that the IPMC is too large to act with consistency and efficiency. The IPMC is kind of like the ASF Membership. The ASF Members list, for those unfortunate enough to be on it, occasionally blows up with problems for which there is no single, perfect solution. The IPMC list does just the same. The rest of the time everything runs smoothly. The way the foundation survives these episodes is to have an elected board that is responsible for navigating through those situations. They don't seek the one single right answer (because there isn't one), instead they listen to a set of options, each flawed in some way an they pick the one deemed to be least flawed from the foundations point of view. The board breaks deadlocks when they occur, the rest of the time the board does nothing more than provide sufficient oversight to ensure the PMCs can operate as Apache PMCs (it gets out of the way). The IPMC has no equivalent. So when it hits a problem with multiple potential outcomes - all flawed, it ends up in deadlock. Look back of the archives, this is happening increasingly often as the IPMC grows. We need a way to efficiently break deadlocks. Let me ask a question... Why shouldn't the IPMC create an equivalent to the one item in the above governance structure that is missing today. That is why shouldn't it have an equivalent of "ASF Members elect a board". It would something like IPMC elect 9-15 Shepherds. These Shepherds are responsible for ensuring that the IPMC membership is heard and that decisions are made for the good of the IPMC. They approve membership of the IPMC, they approve project entry/graduation/retirement but, and this is critical, they report to the IPMC. Most of the time their role is one of delegation to the PPMCs, occasionally their role is to break a deadlock by listening to the IPMC and making the best decision it can. This need not change any other line in the existing governance. It need not change the IPMC relationship with the board. Mentors will still be IPMC members, with binding votes, etc. Since the Shepherds are accountable to the IPMC they must seek to do the right thing, or they will be replaced. Just as the Board can be replaced at any time if the Membership so desires. Of course, we could argue that the IPMC chair already has the authority to do all this. Indeed they do. However, expecting one individual to keep track of all the activity in our podlings is unreasonable. Furthermore, it is harder for one individual to make the hard decisions and suffer the mud slinging from those that don't like the outcome. Just a thought.... of course, my solution is as flawed as anyone elses and I look to the IPMC Chair to find the "good enough" solution that will allow us to move on (sorry Benson). Ross On 27 March 2013 11:55, Benson Margulies <bimargul...@gmail.com> wrote: > I suppose that as chair I ought to be heard from here. I've been off for > Passover for a bit. > > In my view, the IPMC manifests two problems. I'd like to label them as > 'operational' and 'decision-making'. This thread is about decision-making, > but with some people seeing using terms like 'disfunctional', I think it's > important to keep 'function' in context. > > Operationally, we 'started' 1.3 years ago with an acute problem of > under-supervised and/or 'malingering' podlings. Under Jukka's leadership, > we made a series of incremental changes that have considerably improved the > situation. On the other hand, the recent influx of many new podlings > worries me, because 'improved' is not the same as 'fixed'. And I'm not > entirely sure that 'fixed' is possible. I'd like to see us find more > incremental changes that help further, and I'd like them to scale via some > mechanism other than my own personal time. I see this as a reason to put > more thought into shepherds and champions. But I don't see this situation > as 'disfunctional'. > > On the decision-making front, recent phenomena have demonstrated to me that > this group is not succeeding in applying consensus process to decision > making. I could write five paragraphs on what that process is and what it > requires, but I'm not inclined to. I support the proposal here to apply > majority rules to IPMC membership. When consensus process fails here, we > have endless email threads. Many of us find these stressful, > time-consuming, and disheartening. > > Under the proposal at hand, we'd still DISCUSS, and I'd hope that we would > all try to be thoughtful and constructive and look for ways to agree. > However, after a certain amount of discussion, there would be a vote, and > that would be that. > > If this 'works' -- if people here find that it strikes a good balance > between seeking consensus and limiting time and stress, we're good. > > It might not work. Or it might 'work', but some might feel that this large, > diffuse, group, operating by majority rules is either inconsistent with > Apache policy or a bad example for the podlings. In which case someone > might want to dust off the proposals from 1.3 years ago that offered more > or less radical alternatives. I'm personally not ready to go there yet. > > > > > > > > > > > > On Wed, Mar 27, 2013 at 7:03 AM, Bertrand Delacretaz < > bdelacre...@apache.org > > wrote: > > > On Wed, Mar 27, 2013 at 11:44 AM, Justin Mclean <justinmcl...@gmail.com> > > wrote: > > > ...As an aside it seems (and please correct me if I'm mistaken) in > order > > to become > > > a IPMC member you first need to be an Apache member (see bottom of > > [1])... > > > > you don't - Apache members can become IPMC members just by asking, but > > others can also be elected as incubator PMC members. We do have some > > such mentors currently. > > > > -Bertrand > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: general-unsubscr...@incubator.apache.org > > For additional commands, e-mail: general-h...@incubator.apache.org > > > > > -- Ross Gardler (@rgardler) Programme Leader (Open Development) OpenDirective http://opendirective.com
