In Russell Robert’s (of CafeHayek.com) January 19th EconTalk podcast with Eric Raymond, author of the The Cathedral and the Bazaar, Raymond discusses five special circumstances where open source tends to be more successful:
1. Cheap capital goods.
2. Human creativity and attention is the limiting factor.
3. The work is intrinsically rewarding.
4. There’s an objective metric for success.
5. Low communication costs.
I think this is interesting, but I’m not convinced about number 4. Open source can be successful whether the metric for success is objective or not. It can work even better than “closed source” when the metric is not so objective.
Raymond explains open source works well with software development because the metric is clear, “it either runs or it crashes.” He says it doesn’t work so well with something like Wikipedia because there’s “no metric for truth” of an entry “except the judgment of other human beings who may be as eccentric and crazed as you are.”
In his Wikipedia example, Raymond stumbled into the reason why closed source isn’t very good. In closed source it’s too easy to pick the wrong metric for success. Wikipedia’s metric for success isn’t the truth of an entry. The right metric is whether people use it. They do. Jimmy Wales, founder of Wikipedia, reported that 280 million people per month use it on the latest EconTalk podcast. That says that given all the trade-offs, Wikipedia entries provide good enough value for people to use it. Of course, the truth of entries plays a part. Few would use Wikipedia if it were very unreliable. But, it’s reliable enough for many needs.
I suggest reversing Raymond’s fourth special circumstance to say that closed systems work pretty well when the best metric for success is objective and known. I’ve been asked why I think our military, a government ran, closed source organization, is so effective. It’s because the objective is usually clear: to win. Failure is clear as well. The reason why education isn’t a very good when run by the government is that the metric for success isn’t clear.
Even with software the metric for success isn’t as clear as we might believe. In hindsight it may seem so, but it’s not as clear as we think ahead of time. A base metric for success, such as whether it crashes, may be objective. But one solution may meet not crash in 300 lines of code while another does it in 10,000. I’ll take the “300” if it does the same thing.
It was a great podcast. Highly recommended, as are all EconTalk podcasts.