Carr claims that the bazaar model is most and only useful for actions that are divisible and require little coordination effort. Debugging is a typical tasks that can be parallelized. Like in an easter egg hunt, the bazaar model works best if there is a simple task 'find the easter eggs' and there are many independent participants.
The cathedral method is preferred for other tasks, e.g. product development from the ground up. It would be very challenging to develop a product in this way with so many participants. It should be noted, however, that even in the Linux project there was a tight advisory crowd that coordinated the debugging efforts of all the people. This ensured that the overall project moved into the right direction forward. As a side note, Wikipedia, another bazaar style project, does not have this kind of authority which makes it prone to prone to delays and errors.
In short, peer production works best for:
1. Routine tasks that can be pursued simultaneously by a big crowd
2. Projects in which labor is donated and can be parallelized.
3. Projects that are not totally egalitarian or democratic.
Source: Nicholas G. Carr, The Ignorance of Crowds, Business & Strategy, Summer 2007.
