Openness, Open Source and the Veil of Ignorance

Open source collaborations are increasingly among commercial rms whose interest is pro t. Why would pro t-motivated rms voluntarily share code? One reason is that cost reductions can outweigh increases in rivalry. This is especially persuasive when the contributors make complementary products. However, cost reductions do not explain why open source is a more pro table way of sharing than other forms of licensing. Why would rms use an in exible contract like the GPL? I present a model that shows how open source licensing can lead to higher industrywide pro t than would result if a rst innovator could choose the most pro table license once it nds itself in the position of rst innovator. From behind a veil of ignorance, that is, not knowing which rm will be rst, open source licensing creates higher expected pro t for the industry as a whole, and thus for each rm, than if rst innovators were allowed to choose. In the 1990's, open-source collaborations emerged as a new way of organizing software development (Eric S. Raymond, 1999). In an open-source collaboration, members disclose their code so that others can improve it. This is done under various licensing arrangements, for example, a \general public license" (GPL) that grants others the right to use the code in return for a similar right attached to any derivative work of their own. Generally, no money changes hands between contributors. The open source movement evolved in the one industrial context where openness is not required by intellectual property law.2 Nevertheless, openness itself cannot be the driving force behind the open source movement. This is because openness can be achieved in many ways other than the GPL, for example, with proprietary licenses, or licenses that are even more permissive than the GPL, such as the BSD license.3 I thank Steve Maurer, Joachim Henkel and Sebastian von Englehardt for useful comments. I thank the NSF, grant SES 08-0830186 for nancial support. This is emphasized by Maurer and Scotchmer (2006). Patent practice has evolved so that very little about the nature of a program must be disclosed in a patent; see Lemley et al. 2002 at 204-205. For copyrighted source code, there is an explicit exemption. See U.S. Copyright Circular 61. The anomaly is interesting in its own right. It reveals that the theory behind disclosure is a little shaky. The Berkeley Software Development license relieves the user of any nancial obligations, and unlike the GPL, does not require a reciprocal promise to do the same.