Virtual Decentralized Networks: Linux's Organization
barries writes: "Here is an interesting take on the Linux Project which tries to put it in a historical perspective and explain why traditional structures and theory don't fully apply to it. It overlooks a few things but gets most of the basics right." You might want to skip ahead a bit in the paper to get to the Linux-specific sections.
It defies Brook's Law because of its parallel release structure, extreme modularity, "trusted lieutenants" structure and as a consequence, co-ordination costs are almost negligible.
IIRC, Brook's law applies for networked communication. Hierarchical communication is therefore created to reduce this overhead. Extreme modularity doesn't prevent the merging cost. The interesting note here is that this paper doesn't address who is behind the scene to put all of this together.
I am not a Linux developer, but I believe Linus is the ruler of all of them. If he doesn't like the way things integrated, he just demote the component. Thus the maintainer does the job to comply. Recall on how many "new" components, such as ext3 and others must wait and comply to this rule? The rule is pretty rigid and widely accepted by all developers. All his "lieutenants", such as Alan Cox, also applies the same rule.
Moreover, Linux developers are all dedicated people, not just people that wants money to do their work. They want recognition. So, they will try very hard to push their "product" into the development line.
It's not just the Brook's Law, I suppose... Anyway, that's my 2c.
--
Error 500: Internal sig error
Things like this can really distract a reader, ya know?
I don't know why, 'cause it IS actually true...
also, the official name for the Debian Linux distribution is Debian GNU/Linux.
Advanced users are users too!
"The Cathedral and the Bazaar" and the "Halloween Memo" have covered this ground, and better.
There are major unanswered questions to ask about "open source" as a process, but this paper doesn't ask them.