Last week, at the O'Reilly Open Source Convention, Michael Tiemann -- formerly Red Hat's CTO and now vice president of open source affairs -- spoke about the role of Fedora, Red Hat's free Linux distribution. To refute the claim that Fedora represents a fork of its core product, Tiemann appealed to a notion that is best summed up in a phrase popularized by Tim O'Reilly: "the architecture of participation." To meet the needs of the enterprise customers who pay Red Hat's bills, Tiemann said, it was necessary to slow the release cycle and create "a massively long release runway on which Oracle, and Veritas, and BEA, and all these other guys could actually land." But the solution to one business problem created another. It disenfranchised the people in the open source community whose energy and ideas created Linux and continue to drive its evolution. Fedora's goal, Tiemann said, is to be a bridge to that community and to convey both quality and innovation into the enterprise product.
To show how the open source process can yield superior software quality, Tiemann cited a study of contributors to the Apache project. There were 388 contributors overall, but just 15 of them accounted for more than 80 percent of the code changes. At what point on a commercial project, Tiemann asked, is the incremental cost of adding another developer negated by the diminishing return on that investment? No matter where on the curve you make that decision, it limits the quality of the result. With open source, though, nobody gets shut out. "Developer No. 388 was not drawing a high salary," Tiemann said, "twiddling his thumbs waiting to ding the tambourine at the end of the symphony with his one bug fix, but rather was a Tomcat developer, or a Mozilla developer, someone who -- in order to get his job done -- needed to fix this one stupid bug in Apache and move on."
It's one thing to talk about quality and another to talk about innovation. Even if you agree that the series of fixes culminating with Developer No. 388's patch really does deliver a level of quality that the closed-source model can't economically match, you're not likely to regard Developer No. 388 as an innovator.
To showcase Fedora's role as Red Hat's "innovation platform," however, Tiemann chose an odd example: SELinux (Security Enhanced Linux). Similar to the trusted versions of Solaris -- AIX and IRIX -- SELinux implements a feature called mandatory access control, which works with a set of security policies to protect objects independently of the permissions assigned, or not assigned, by their owners. By including SELinux in Fedora, Red Hat hopes to get developers to write applications that rely on it. Of course, SELinux did not emerge from the open source community. Quite the contrary, it's a research project of the National Security Agency.
Discussions about open source and innovation tend to cluster around two opposing memes. One says that open source can't innovate; the other that only open source can innovate. Both are wrong. Sometimes large, well-funded R&D programs can achieve breakthroughs that lone geniuses can't. And sometimes the reverse is true. Either way, the real innovation of the open source movement is the architecture of participation. It can help turn a good idea -- wherever it came from -- into a best-quality implementation. Software companies that don't choose the open source model have to find other ways to recruit and reward participants.