Pragmatic Agile Weblog

Agile software development in real life

Funambol Code Sniper v2: agile open source development?

Ok, at Funambol we like challenges 🙂 Who knows about Funambol development, may know about the Code Sniper program. It is a program that with the aim of encouraging open source development around the Funambol platform, offers bounties to develop projects that the community would like to have.
The original program worked fairly well, but shown also some limitations. The main issues we have been facing are:

  • technical specifications vs features specifications: the first thing a developer was requested to do was to write a specification document so that the community could understand what was being developed. Typically, this was more a technical document that explained how things would have been done. This may be a good thing, but it does not show which value a potential user would get from the software.
  • whole or nothing: being the project based on a monolithic specification, the approach of the development was in many case a monolithic effort. The developer started the development one her own, and until everything was done, the community did not see much. In other words, this approach does not encourage an iterative and incremental development so that potential users cannot see anything until the entire project is done. At that time, feedback may become not only not welcome, but even frustrating (you tell me know that this is not what you wanted???).
  • long development cycle: the above, usually turns also in a longer development cycle that results in a higher risk that the developer does not complete the work for lack of time, abandoning the project in an unknown state. This makes it more difficult for another developer to take over.
  • difficult work recognition: how to recognize (and then remunerate) the work done? in a whole or nothing approach we are forced to recognize the work done only when all the work has been completed. But this can be unfair if a developer has done some work, but could not complete the project. Plus, it becomes a barrier to picking up a sniper by the developer community.

In short, the above limitations result in a higher barrier to the participation to the program. Stefano Maffulli tackle the problem and redesigned the program to make participation simpler and wider, even for non-developers, and to provide long term incentives. Code Sniper v2 rewards not only developers but also testers and users that provide feedback. It is simpler because it only consists of only three steps:

  1. Identify the target – pick from a list of projects, which are ideas provided by Funambol or the community
  2. Contribute – write feature requests, develop code, test or give feedback. Development is carried out by volunteers that pick items to work on from the list of feature requests or from bug reports. Each time an item is completed, a new version is released
  3. Collect your reward – for each released version, contributors collect the reward and the cycle can start again, providing a long term incentive to collaborate

You can find more information here.
There is an interesting aspect of the new Code Sniper program that I want to highlight. The new program is designed around principles that come from the agile community. Let’s see how, in addressing the challenges identified above.

  • technical specifications vs feature specifications: the new program encourages the use of User Stories to describe requirements. This changes completely the approach to how specify what is going to be developed in two ways: first, they are written with the point of view of a user, describing what a particular user will have instead of focusing on how something should be done or should work; secondly, user stories encourage the description of features in smaller pieces, more than in a monolithic set of functionality. You can see an example in how to write a Funambol Sync Client, the agile way.
  • whole or nothing: being able to break features in small user stories, each bringing value to the user avoids the whole or nothing approach. It also allows to see the development of a component as an iterative process. At any given time, the application does something. User story after user story, the application builds up, giving more value to the user at any addition. This encourages an earlier adoption of the software. Not all users will be satisfied at any moment, because maybe the application does not give enough value to them. But other users or developers can benefit from even an incomplete application, thus being more stimulated to contribute.
  • long development cycle: each user story must be a self contained feature. It is not allowed to check in a user story until it is complete and working. The goal is to have a working application at any given time; maybe incomplete, but what is in must work. This allows to break the development in smaller cycles, to release more frequently (potentially at any time), to be able to test a feature earlier. This should also stimulate participation because a developer knows that she does not necessarily has to be in charge of a big task (and therefore a big effort), but she can pick up the user story that she can do in the time she has available.
  • difficult work recognition: now that the work is divided in user stories, it is also much easier to recognize the work of a developer. A developer has done when a user story is working. This also means that more activities than just development can be recognized. For example the definition of the user story itself and the acceptance criteria to consider it done; also the testing to accept or the user story can be rewarded. For example a user story could be picked up by two people. One (the final user) would scope the user story, define the acceptance tests and perform the testing. The other person could develop the code.

I believe this is a kind of ambitious goal and truly experimental. We do not hide that there is the other side of the coin. Probably, being inspired from agile methodologies, this approach requires more educated people. But hopefully, as agile is emerging as the most popular development model, this will not be a big issue. Anyway, in the spirit of agile, we will try, inspect and correct!

Feel free to leave your comments and feedback!

Advertisements

December 18, 2009 Posted by | Agile, Funambol | Leave a comment

Informative workspace – the scrum kiosk

One of the biggest challenges you have in a organization with 6 scrum teams with some people remote and different stakeholders introduced at different levels in the scrum methodology, is sharing some significant information about the status of the development.

Each scrum team tends to create its own working style, a sort of ecosystem that best fit people in the team. The challenge is to encourage that everyone, inside the scrum teams, but also external stakeholders, can share the same big picture.

One way to encourage better mindfulness amongst the entire team is to create an “informative workspace”, something that tunes anyone in to the status of the project.

Better than  using my words to explain it, I will borrow an excerpt of James Shore & Shane Warden’s book The Art of Agile Development.

Your workspace is the cockpit of your development effort. Just as a pilot surrounds himself with information necessary to fly a plane, arrange your workspace with information necessary to steer your project: create an informative workspace.

An informative workspace broadcasts information into the room. When people take a break, they will sometimes wander over and stare at the information surrounding them. Sometimes, that brief zone-out will result in a aha moment of discovery.

One recent improvement we have done to make our environment more informative is what we called the scrum kiosk.

We installed a monitor on the wall and we slide-show projects facts: deadlines, status reports, process information and reminders, burn down charts and so on.

It’s a live snapshot of the current iteration and release.

We’ve implemented it in few simple steps using the following software:

DejaClick: Firefox plugin for web browsing recording. It automatically reproduces the main paths done browsing our internal systems and wikka pages

– a Fullscreen Firefox plugin

– a simple Java app to put everything in an infinite loop

You can see some pictures below.

In addition, since we have remote people, we plan to stream the content of the monitor to our internal network, so that everyone can see what’s displayed. Indeed it is not the best solution, but it is something in meanwhile we find a better idea.

Any comments or suggestions is welcome!

December 2, 2009 Posted by | Agile, Funambol, SCRUM | 3 Comments