From the corporate world I frequently hear how hard it is to predict and track what upstream developers do. On the other side, developers that work part or full time upstream frequently underestimate the need for communicating what they do in a way that enable others (or themselves) to provide deadlines and effort estimations. Upstream and product "time lines" and cultures often differ too much to be compatible under the same environment.
Developers that come from the "product" side of the story often refer to agile methodologies like Scrum as a good approach to solve this collision. It is unclear though the applicability of this and similar methodologies to highly distributed environments, where latency is high. Although most accept today that FLOSS development is agile, I find hard to assume that we are close to find good answers when it comes to development methodologies and upstream.
There are several different scenarios I have worked on where I have to question everything I knew up until then about this particular topic. In a couple of occasions, with a bigger or smaller community, the people I directly worked with were upstream. This allowed us to define up to a great extend the methodology we worked with.
In Linaro now this is the case for a particular project. But in general, in my Group we work upstream (in the Linux Kernel), that is, we cannot define the methodology since it comes defined by the project itself. It is not the first time I face this situation although I never did at this scale.
When you are upstream, the methodology used by the core team, together with how the project work packages and workloads are organised heavily influences the success in getting contributions from third parties. Together with other key variables, how you manage latency determines how many people can follow you and potentially contribute. In summary, the faster you move the harder it becomes for contributors to collaborate with you.
Hence agile methodologies come with a high risk: isolation. And this is obvious by just analyzing the prerequisites associated with the methodology itself.
Obviously this does not mean that scrum and other methodologies are not great ones. I am just pointing challenges associated with apply them when working in a community that you drive. You need to balance the efficiency of your team with the contributions you might potentially loose from externals.
The challenge is even bigger when you are just a participant in a community. In some way, your team goes beyond your colleagues at work. Latency is frequently too high to even consider yourself and the people that works close to you "a team" in the classical sense of the word.
If your own team is distributed, even if they are in the same time zone, then we are talking about the real deal, the master challenge for managers and developers.
Imagine now that your team is formed by 10 people distributed in 6 different time zones in 3 different continents. Then you take one of those cool books that tries to explain how to create great software using this or that methodology and, in order for you to finish it, the writer must be a really good one or you really need to be an avid reader :-)
Still you need to make plans, to predict when a feature will be finished, when it will be merged, you still need to manage dependencies, expectations, budgets.... you still are tight to somebody else needs and requirements. You still need to fulfil expectations despite the fact your control over the environment is limited. You have team mates that depend on your work, that needs to know what you are doing and how..... you are still part of a business no matter if you develop upstream or not.
How can you make compatible the upstream development processes used in the community you are contributing to with the way your company works? And how about your customers and partners, if you have direct relation with them? How can you take the good things that agile development propose and apply them to an environment with high latency, where one of your bigger challenges is to have an efficient team meeting, since people are distributed across the planet?
At Linaro, the engineering management team, together with our PMO, are taking a close look at this challenge, in order to iterate from our current system to a more efficient one, better adapted to our new reality, after our significant growth during the past 18 months. The idea is to find a place where corporate and community processes meet... without colliding. You cannot stop adapting, innovating, trying new things..... or you pay a high price.
One of the great things about Linaro is that we are a unique environment so we can innovate at various levels, not just at the technology one.
Open Sorce Forum at CebIT
For those of you attending to CebIT, let me tell you that I am giving a talk about what we do at Linaro on Monday March 16th at the Open Source Forum. It is the first time I talk about my current job in public so I am very excited about it. It will be also my first time in CebIT so double doses of excitement. And yes, I will talk a little bit about Linaro new initiative, 96boards.
Core Dump
The past months we have published a few articles about what we do at Core Development Group at Linaro. Our blog is called Core Dump. Check it out.
