Phys. Rev. Lett. 101, 218701 (2008)

In the 1940s, linguist George Kingsley Zipf found that the probability distribution of a wide range of variables, including word-use frequency and demographic distributions, depends on the rank of the variable according to a power law. Now Thomas Maillart and his colleagues at ETH Zurich in Switzerland report empirical evidence from the spread of open-source software that an explanation posited in 1955 is correct.

This came from economist Herbert Simon. He thought that Zipf's law stems from the growth of a population of which the size varies at random but with a standard deviation proportional to that size. An analysis of exceptionally detailed data reveals that the Zipf law in incoming links to packages of the Debian Linux computer-operating system is supported by exactly this growth process.