The Apache Software Foundation Blog

Monday February 06, 2017

Success at Apache: Asynchronous Decision Making

by Bertrand Delacretaz

Asynchronous decision making is a key enabler of our geographically and culturally distributed Open source teams. In this post I'll explain the ingredients that make it work at the ASF.

I became active in the ASF in 2001 via Gianugo Rabellino - he was the one who started the discussions with Apache Fop about me donating the jfor XLS-FO to RTF converter that I had developed earlier. It was already too late to uninvent RTF which is a terrible format, but I digress. I am currently a member of the Board of Directors of the ASF and have been doing a lot of thinking (and presentations) about what makes the ASF tick in terms of collaboration and Shared Neurons.

If synchronous decision-making meetings were required in ASF projects, even using remote channels like IRC or videoconference, we would move forward at a snail-like pace, as just finding a time where all stakeholders are available is almost impossible in an environment that has no managers and no central schedule.

Meetings are also very expensive when you are working on a maker's schedule, as described by Paul Graham [1]. Frequent meetings ruin the productivity of craftsmen, and there's lots of craftmanship in our industry, especially when you're building leading edge stuff.

So, what's needed to enable people to make collective decisions asynchronously, without requiring meetings?

The first thing you need is a central asynchronous communications channel. Which technology you use for that doesn't really matter, but it has to allow everybody to get the same information, and provide a usable way of having threaded discussions, where you can branch off on a topic while ignoring other topics being discussed on the same channel. This can be as simple as a whiteboard if people often visit the same place, or as elaborate as web-based forums, accessible from any mobile device so you can bother^H^H^H^H^H reach people everywhere. At the ASF we use plain mailing lists for that, very successfully when people use them with the right discipline (see the appendix below). Archiving this channel is very useful, to allow newcomers to get a feel for how things work as well as documenting the reasoning that led to each decision and avoid having to repeat things over and over.

The second required tool is a way to build consensus, where you avoid deadlocks and make sure decisions go forward. Unanimity in decisions is ideal of course, but the second best is consensus, defined as widespread agreement among people who have decision power. Requiring unanimity or allowing vetoes in decisions can block progress, so at the ASF vetoes only apply to a very limited set of decisions types, as defined by our voting rules [4]. In companies, decision power can be based on hierarchy to break deadlocks. That has to happen sometimes, but abusing it can cause employees to lose their autonomy and purpose, which kills your team in the long term.

To keep track of each decision, a case management system is ideal. You could work without that, depending on your team's size and the number of decisions that you take, but it's very convenient to be able to discuss the details of a given decision and keep associated information in a single place. You don't need complex software for that, at the ASF we use fairly simple issue trackers. Those are Web-based systems where each case is handled on a single page, with a history of comments and actions. Some non-urgent or very hard decisions can take a long time to reach closure, and it's very useful to keep their history in a single place, if only to avoid having to explain them again to new members of the team. In a low tech environment you could just use a single sheet of paper to briefly document each decision with the key points that led to it, and keep those in binders or physical files.

A nice side effect of using case management software is that each decision gets a simple unique identifier, like FOO-123 for the 123th ticket of the FOO project. This removes any ambiguity as to which issue one's discussing, by mentioning those identifiers in conversations.

So, in summary, the following should allow your group to make decisions asynchronously, without requiring meeting and with a written trace of everything that happens:

  • An archived asynchronous communications channel, where everybody can get the same information and threaded discussions can take place.
  • A way of building consensus, including fair rules for breaking deadlocks.
  • If possible, a case management system to keep track of each decision's details, in a much cleaner way than the often messy discussions that happen on the asynchronous channel. 

Semi-asynchronous decision making at the ASF

I've been a member of the ASF's Board of Directors for a few terms now and I'm still impressed by how efficient our monthly phone conferences are. The meeting regularly lasts only 60 to 90 minutes, during which we approve around 50 project reports, vote on a few resolutions and often address a few discussion items.

Besides a few simple things like good phone discipline and a side channel for less important comments (and jokes), the main reason this meeting is so efficient is that almost everything is decided in advance.

Board members are expected to read the project reports before the meeting, and a dead simple case management system (described below) helps discuss issues in advance, and find out which reports require a more extensive discussion.

Assuming the majority of board members have read the reports in advance, and flagged them as ok or requiring discussion, we don't need any housekeeping time during the meeting. Everybody shows up with a clear view of where difficult discussions might arise, so they have time to prepare for that, including asking others for clarification before the meeting so we can resolve any outstanding issues without delay.

The case management system that we use for this is extremely simple, but in terms of enabling asynchronous (or rather semi-asynchronous) decision making it fullfills its role. Our meeting agenda consists of a single text file in our source control system, with a simple structure that provides for a small discussion space for each report that we have to approve and each resolution that we need to vote on.

The agenda file structure looks roughly like this:

Call to order
Roll call
Officer reports
Project reports, headers and discussion space
Board Resolutions with discussion space
Appendix: Full Project reports and other supporting material

And a project report header and discussion space is as simple as this:

E. Apache Blazinator Project [Bob Blazer / Bertrand]
  See Attachment E
  [ Blazinator.
    approved: bd, mm, dd, db, jc, ldv
      bd:  Not sure why LEGAL-123 blocks their release
      ldv: They are waiting for the committer to supply
           an updated iCLA as the received one was 
      bd:  Ok, thanks, approving the report then.

This simple block of structured text builds a very simple "case management system" for the case of approving the Blazinator report.

The "approved" line indicates which board members have approved the report, on a single line so that simple text-based tools can validate and count the approvals.

The "comments" section allows stakeholders to comment on the report (which is found in an appendix later in the text file), and reply to each other's comments to hopefully reach closure before the meeting. If this happens, approving this report takes almost no time in the meeting, the chairman can just list the project names ("case identifiers" according to the above terminology) of such pre-approved reports, asking if anybody's opposed to approving them.

Combined with the ASF board's mailing list, this builds a very simple and very efficient system for semi-asynchronous decision making. Most decisions are taken before the meeting, and the participants can spend their time where it actually adds value as opposed to exchanging boring status information during the meeting.

Try it yourself!

Many ASF and other Open Source projects release world-changing software while having no or very few meetings, demonstrating that these techniques work.

If you're bogged down with inefficient or useless meetings, I suggest that you try applying these principles to a meaningful subset of your decision making activities. People will need to hone their skills to work efficiently in this way, but the rewards can be huge for distributed teams.

Appendix: Mailing lists at the ASF

At the ASF we use mailing lists as our central asynchronous communications channel, based on our if it didn't happen on the dev mailing list, it didn't happen rule [2]. Mailing lists might be seen as tools of the past when you compare them with the latest shiny tools, but they remain a ubiquitous way of communicating in loosely coupled remote groups, especially when used with a strong discipline of Precise Quoting [3] and paying attention to meaningful subject lines. Unfortunately I hear some "modern" email clients make a mess of that quoting - just stay away from them.


[1] - Paul Graham, Maker's Schedule, Manager's Schedule, July 2009

[2] - The Apache Project Maturity Model, ASF community development team, 2015.

[3] - Gianugo Rabellino "[OT/Rant] Quoting", message to the cocoon-dev mailing list, January 2002

[4] - ASF voting rules, created in 1999 probably, or even earlier among the Apache Group.

 # # #

"Success at Apache" is a new monthly blog series that focuses on the processes behind why The Apache Software Foundation (ASF) "just works". First article: Project Independence Second article: "All Carrot and No Stick"

Monday January 09, 2017

Success at Apache: "All Carrot and No Stick"

By Danny Angus

When the ASF launched their "Success at Apache" series I offered to share my own experiences. If you read on, remember that this is my personal experience and that others may disagree with me, but as you'll see, that's really part of the fun. 

For a bit background I’m currently the Project Management Committee (PMC) Chair of Apache Labs and in my day job I’m a "Divisional CTO" for a FTSE250 technology company. I first came to the ASF around 2000 when I was part of a startup - I was a CTO then too, it was the dot com boom, and it was just me and a couple of guys. We were considering a partnership with some researchers who wanted to commercialise their work, and were looking for a bit of software that we could use as the foundation for a product because a) we couldn’t afford to write it or buy it, and b) we didn’t have the knowledge anyway. What I found was Apache James , so I downloaded it, got it up and running, and did some prototyping, but we quickly realised that it needed work if we were going to be able to use it in production. I dug into it a little, subscribed to the mailing lists, asked questions and figured out what needed to be done to fix and extend what was already there, then started to modify it locally. Meantime I found myself answering other users’ questions on the user list, and one day noticed that I was actually answering more questions that I was asking. Shortly after that, that I was answering more questions than anyone else. Then I started submitting patches to the developer list (this was in the days of CVS: long before git!), which were reviewed and committed for me by the committers … but eventually they got bored with that and decided to extend commit privileges to me so I could do it all myself.

My experience illustrates an important characteristic of Apache projects: the fact that you can just turn up and get involved. Another very other important characteristic is that we are a meritocracy: demonstrating your capability is all you need to do in order to gain more responsibility; demonstrating your willingness and trustworthiness should be enough to get you the job. "Karma" is a word that is used to mean "access permission" in many Open Source projects, and we used to say that if you knew how to ask for karma properly, that was itself a sign that you could be trusted with it. Of course we were a much different organisation in those days, but the principles of a community built on merit and trust are still core to our identity. It's no coincidence that organisations cannot be part of our community: only individuals. Organisations are an important part of the world in which we exist, but we don't exist for their benefit, we only exist at all because as individuals we each bother to turn up and do stuff, from the guy who one time downloads and installs the Apache HTTP Web Server to Sam Ruby, our current (and can I just say excellent) President, everyone is contributing in their own way to the life of Apache and achieving benefits suited to their own, personal, motivations. So it was OK for me to focus on my own and my employer's priorities, which meant that I could learn from my new friends, develop the software we needed at work and become part of this amazing community all at the same time.

My experience of Apache is that it is what I would call "all carrot and no stick". I think that is the most healthy model of Open Source, as it is predicated on the fact that every participant will benefit from their participation without the need to contribute more than they are prepared to do. For me, focusing my contribution on the things I knew about was not only the most efficient use of my time, in terms of meeting our company's product goals, but it also allowed me to learn from others who had, and continue to have, way more knowledge and experience than I, and to benefit from their work. Mixing with these amazing people, many of whom are now real friends of mine, has taught me more than I would ever have learned any other way.

At this point in my involvement Apache went through a bit of what has diplomatically been described as "navel gazing", and settled on the idea that the organisational structure should be very very flat, and there should be no limit to our growth. As long as our standards were met by projects and people, we would welcome them both into our community. Those standards are partly about merit, partly about legal protection, one of the key roles Apache plays is to provide a degree of protection to projects and the people contributing to them, from the threat of bullies, trolls, and gorillas with expensive lawyers; and partly about ensuring that the behaviours and practices that define our identity and have contributed to the survival and the success of our organisation are continued by new generations of people in new projects using and creating technologies that we could hardly have dreamed about 16 years ago.

Before long the dust settled and I found myself voted to chair the Apache James Project , which was a whole new dimension of interesting. Chairing a project using only positive motivation teaches you a lot about people, including yourself, and I have a few observations about successful collaboration that I have found to be helpful both at work, where I strive to implement bottom-up decision making, and at the ASF where I want to make a positive contribution and see our communities flourish:

  • Free your mind.The collective sense of direction may not be what you expect, there have been times when I have been very sceptical about the reality of great sounding ideas, but I have also learned that it’s OK to go down the wrong road because most of the time it makes little difference in the end, usually you learn a lot regardless, and if people are really behind it you stand a much better chance of success than if the really good idea has all the fun of a death march. One phrase which is often used to summarise the spirit of Apache is “Community over code”, put the community first, and the code will follow.

  • Listen, and be supportive. There are a lot of different people involved in our projects with a lot of very different motivations. They are mostly all valid, and mostly all equally important if that even has an absolute scale. There are students studying our code, asking questions using our software and maybe fixing defects so that they can learn, there are employees of corporations who are being paid to protect their investment, to implement the product roadmap and maintain some predictable velocity, there are researchers who are pushing the boundaries of their chosen topic, there are people whose livelihood and success depends on a project, and those who are involved because it is a release from the pressure of things with names like "impact", "benefits", "deadlines" and "goals". Moderate or steer the discussion to ensure that all sides are heard, a meritocracy needs to listen to everyone not just the most vocal or assertive, and when I say listen that doesn’t mean formulating your own response while someone else is talking. Support people who you agree with, help to realise other people’s ideas, collaboration is only achieved by being truly committed to each other’s success, not just your own.

  • "A's hire A's B's hire C's". Find, support, and mentor the next generation, when your success depends upon the community it makes sense for you to put some effort into creating the best community you can.

  • Use Positive Language. When I was a kid being mean to my sister, adults used to say, "If you don't have anything nice to say, don't say anything at all". That's great advice if you’re involved in any collaborative venture, but doubly so when it is something like an Open Source project where you are usually communicating using written English, with people you don't know well, who might not have the same language skills as you do, who live in a different time zone and sometimes have very different cultural background than you. On top of all that you"re often debating the details of highly abstract technical concepts. The communication barrier itself can cause a kind of baseline of frustration so go easy on the negativity, one thing I like to do when I strongly disagree with someone is to write how I feel, then try to reword it using only positive language, it might sound like touchy-feely hippy nonsense to you, but you will be surprised how effective changing "I think you’re wrong and here’s why..." into "You have clearly thought a lot about this, I wonder if you have considered...". Alienating people is not the way to get your point across.

  • Learn to be a good loser. You don't own your projects, not here, and you're not the smartest person here either (OK so that’s not going to be 100% true, but there are 5,938 Committers today which makes it about 99.98%) recognising that and learning to embrace the collective view is hard for some people, but being able to step outside your subjective point of view and make a success of something you didn't believe in is a lesson in leadership that is definitely worth learning, because if not, your growth will be limited by the ideas that come from your own head, not accelerated by other people.

  • We are making it up as we go along . Yes, it sometimes seems from the outside like we have it all sorted and nailed down, and that we want to lawyer up and suck the life out of every fun thing (I mean we have a major software licence with our name on it, how grown up is that for goodness sakes?)  But the truth is that Apache, The Apache Software Foundation is, and probably always will be, a work in progress, hopefully will be at-least-good-enough to survive the next unexpected storm, and there have been several of those already, but the only way we ever find that out is when it hits us. Over a relatively long period we have figured out, adopted, borrowed, adapted, had donations of, and thunk out with nothing but our own brains, a whole load of ideas about effective Open Source collaboration, governance, legal shenanigans, marketing, community building, and so on. Things that work well, some that mostly work, and some that are sometimes rubbish, but better than nothing. We write these things down and propagate this good practice amongst projects because it is the bedrock on which our foundation rests, but that doesn’t mean that it can’t change, we correct, adapt and evolve our best practices all the time, this is how we adapt, this is how we have survived and remained relevant in a field that seems to change almost beyond recognition every four or five years. And, being a meritocracy, if you don’t agree with the way things are, if you think it is out of date or ineffective or pointless, don’t complain, stay and fix it. We have another saying which is that "you can scratch your own itch" - don’t be passive, if you care about it, do it.

    The important point about Apache is not that we have rules and committees but that we have these things because they have been shown to help us do the right thing, to help us to live by our principles and to provide a home for Open Source projects that will equip them to survive amongst the commercial sharks in the ocean of the software industry.

  • Finally: Define your own achievements. Whether you are doing it because you need some software, or because, like me, you just found it and it wasn't quite ready, whether you want to make friends, or to learn something new, whether it is because you are being paid to promote your employer's best interest, because you want to explore new ideas, or because you always wanted to write a book, Success at Apache is yours to define. Create your own measure of success and let us achieve it together.

# # #

"Success at Apache" is a new monthly blog series that focuses on the processes behind why The Apache Software Foundation (ASF) "just works". First article: Project Independence

Monday December 05, 2016

Success at Apache: Project Independence

By Mark Thomas

I've been involved in The Apache Software Foundation (ASF) since 2003. I was using Apache Tomcat at work and I hit a problem that needed a new feature to be implemented. There was already an enhancement request in Bugzilla so I submitted a patch. After some re-work by the project committers, the patch was applied and the feature available in the next release. I enjoy problem solving, so I started to take a look at the other open Tomcat bug reports and my involvement grew from there to include Apache Commons, the Infrastructure Team, the Security Team and, most recently, the Board of Directors to which I was elected in March 2016.

Apache Tomcat has always been at the heart of my involvement and is where I spend most of my time. Tomcat started with a donation to the ASF by Sun in 1999 and, some seven major versions later, the project continues to be very successful. A significant part that success is due to the involvement of a wide range of individuals from different companies. The reason those companies are happy co-operating on Tomcat is because of the importance the ASF places on project independence.

There are many aspects to project independence but, for me, the most important is that committers and Project Management Committee (PMC) members contribute to the project as individuals and do so with the intention of doing what is best for the community as a whole. Some committers contribute in their free time – I did for the first five years or so with Tomcat – and some are allowed /directed to spend time contributing to Apache projects by their employer. However, those committers contributing on their employer's time still need to act in the best interests of the community rather than the best interest of their employer.

To give a specific example, my employer has a product that is built around Apache Tomcat. The sales folks at my employer asked if I could add a feature to this product. The problem was that this feature required access to low-level Tomcat internals in order to implement it effectively. For this to be possible, I would have needed to make some ugly API changes to Tomcat to provide the integration points required. Rather than try and push those changes through, I persuaded my employer that it would better to donate the entire feature to the Apache Tomcat project.

This feature also demonstrates other important elements of a successful ASF project: the ability to make decisions in public and always aiming to achieve community consensus with those decisions. As the development of this new feature progressed, the design evolved as the community reviewed the commits and suggested improvements. This isn't always the quickest way of working but the quality of the end result – both technically but more importantly in terms of community health - more than makes up for that.

The perception of project independence is as important as projects actually being independent. It is a key factor in many projects choosing the ASF as their home so projects need to ensure that the perception agrees with reality.

Things can and do go wrong. With 350 projects it is pretty much a given that there will be a handful of ongoing issues at any given time. For example, there might be an attempt to push a project in a particular direction or to suggest that some external entity controls / leads / manages the project. Typically these are self-corrected by the PMC. Sometimes the PMC needs help to resolve the issue e.g. from V.P. Brand Management or possibly the ASF Board.

Being a board member is often viewed as more significant than it is. I have no more status in Apache Tomcat, Apache Commons or any other project as a board member than I did before my election to the board. I can still have bad ideas and my fellow community members still point it out when it happens. I don't get to always have my way just because I am board member. It is the board as a whole, rather than the individual board members, whose voice carries significant weight. It is fairly rare for any board member to speak on behalf of the board. To give that some context, I've probably done it no more than once a month since joining the board. It is sufficiently rare that board members always include an explicit "on behalf of the board" when speaking for the board rather than as an individual. Sometimes this point isn't appreciated and the views of an individual board member are incorrectly taken to be the views of the board.

The ASF board is also very different to a corporate board. The board manages the Foundation but it is the PMC that manages the project and sets the direction. The board has no role in the technical direction of a project. The board has responsibility for corporate governance, finance, legal etc., but its primary role is monitoring, mentoring and coaching our project communities to help keep them healthy. As part of this, the board reviews all projects on a regular basis. Newly graduated projects are reviewed monthly for typically 3 months before moving to quarterly reviews. The project V.P. (PMC Chair) is an important part of this. They are the eyes and ears of the board. While the board will look for warning signs as part of its regular review, the V.P. has much more in depth knowledge of the project and can flag specific issues early. Where issues are identified, the aim is to get the PMC to self-correct. The board will provide mentoring / coaching / guidance as necessary but it will be the PMC members who do the work to correct the issue.

As an example of the board working with a PMC, earlier this year the V.P. for a particular project became unavailable. The board became concerned because the regular reports were not being produced for the project. In this instance, no-else on the PMC had experience of being a project V.P so the board worked with the PMC to identify a new V.P. and to then mentor the new V.P. as they found their way in their new role.

For the last 17 years, the ASF has provided a home for a large and diverse set of open source projects. Key to this success has been the importance the ASF places on project independence as part of the Apache Way. By continuing to adhere to the principles of the Apache Way, I am confident that the ASF will continue to be successful for another 17 years and a long way beyond.



Hot Blogs (today's hits)

Tag Cloud