Community-based production: do they need a roadmap? The Debian case
The Debian Project yesterday announced the release of “etch”, the last version of Debian.
The press release reported that it took 21 months of development to create this release. Whether you consider contributing to Debian a work or a joy, it would be interesting to know how much would have cost to develop “etch” from scratch.
Roadmap by Pinocure
Being based exclusively on volunteer contributions, Debian can’t grant the availability of all packages included in the previous version, as results from “Evolution of Volunteer Participation in Libre Software Projects: Evidence from Debian“. Packages maintained by volunteers who left the project become unmaintained (“orphaned”) and the probability that an orphaned package gets adopted by other maintainers is not 1.
[..] maintainers who left Debian between July 1996 and December 2004 were responsible for 33.5% of the packages in 2.0, 67.5% of these packages can still be found in 3.0.
The Constitution itself can’t help much when a volunteer decide to exit and no one is willing to take care of his or her tasks. It is worth to notice that within an hybrid production model paid employees are often responsible for less attractive tasks, as results from “GNOME, a case of open source global software development“:
Paid employees are usually responsible for the following tasks: project de- sign and coordination, testing, documentation, and bug fix- ing. These tasks are usually less attractive to volunteers. By taking care of them, the paid employees make sure that the development of GNOME continues at a steady pace.
Corporate production has to be on Time on Budget. The firm solves the problem of finding the efficient management of human resources through time not allowing the free entry and exit, and delegating production control to a manager.
Community-based production on the contrary allows volunteers to enter and choose their tasks. Volunteers choosing what to do apply for tasks they like, and that they are likely to accomplish effectively. They can also freely exit from a project though, or not to end their tasks on time.
How open source firms will approach the hybrid production model? Whatever is your guess, read the following (old) excerpt from the Debian Weekly News – December 2nd, 2003:
Debian Roadmap? The project was asked if there was a roadmap for the Debian distribution, so that certification can be organised accordingly. Ben Collins pointed out that Debian hardly has release goals and Jonathan Dowland added that a smaller group of loose-knit volunteers has managed to agree on a roadmap.
Roberto Galoppini and 
Martin Michlmayr 2:42 pm on May 10, 2007 Permalink
I believe roadmaps are gaining importance in free software development too. IMHO this is related to increased complexity found in many successful projects (both in terms of the size of the development community and the code base itself), which requires a higher degree of planning than in the past. For example, shortly after Debian 4.0 was released, the release managers contacted the maintainers of every large software package in Debian (e.g. the Linux kernel, KDE and GNOME) to obtain more information about the release plans of these projects. This information will be used to create a release plan for Debian. Furthermore, during the development cycle of Debian 4.0, release goals were defined in a much better way than this was done in the past. There was also a split into release blockers and release goals to make it clearer which work is absolutely needed before a release can be made.
In summary, I don’t believe the absence of plans is something inherent with free software development. I believe there will be more planning as more projects gain considerable complexity and size, and to some extent we can see that already.
Roberto Galoppini 5:06 pm on May 13, 2007 Permalink
Martin thank you to join the conversation. I agree with you, the absence of plans is not inherent with free software development.
You mentioned GNOME and other projects where paid developers are on duty for unsexy tasks. Do you believe that the hybrid production model might be the third way?
This way we might get the best of both world, but harmonizing contributions is not straightforward, though.
Martin Michlmayr 12:47 pm on May 30, 2007 Permalink
There’s certainly a trend towards hybrid models, even though they are (or may be) associated with certain problems too. There’s a fairly good paper about the issue of control in the Netbeans community. The question there is who is actually in charge of the project – community or a company (Sun in this case).
Reference:
Jensen, Chris and Scacchi, Walt: Collaboration, Leadership, Control, and Conflict Negotiation in the Netbeans.org Community
Roberto Galoppini 3:07 pm on June 2, 2007 Permalink
Thank you Martin, I didn’t read that paper before.
The way a corporate actor open the development process to others can deeply affect results. Looking at Eclipse vs Netbeans popularity I wonder at which extent it is to be related to the way IBM and Sun backed their respective projects.