Application and Philosophy

All of these licenses have been used in practice, both in licensing software maintained in the open source community and in providing the basis for commercial applications of programs derived from open source models. The BSD, MIT, and Apache Licenses, longer established and more frequently adopted than the Academic Free License, provide the examples described in this section.

Each of these three licenses has contributed to the widespread commercial adoption of the programs they license, frequently (though not always) through incorporation into products distributed under a proprietary license. This is completely consistent with the language and intent of the licenses. This also reflects their place of origin. For example, both Berkeley Unix and the X Window System were research projects; the goal of their creators was to explore technology, to provide a proof-of-concept implementation, and then to permit others to build on that work. Commercial applications readily followed successful implementations of research ideas.

BSD Unix became the basis for commercial versions of Unix ranging from Sun's Solaris to Apple's Mac OS X. BSD-derived proprietary versions of Unix outstripped the commercially licensed AT&T versions relatively quickly, and they dominated the commercial Unix market until the 1990s when Unix was challenged by GPL-licensed Linux distribution. The TCP/IP software stack that was part of the Berkeley networking release became the basis for almost all commercial TCP/IP stacks, including Microsoft's. The X Window System became the standard GUI platform for the Unix workstation market, displacing Sun's proprietary NeWS windowing system. In addition, even as these commercial implementations became available at the same time, open sourced implementations continued to be widely available and accessible for modifications and improvements by programmers.

Despite setbacks from a lawsuit from AT&T that was ultimately settled out of court in 1992, Berkeley Unix still has many million installations, running such well-known sites as Yahoo!, and it continues to be modified and improved. Moreover, and partly as a result, later commercial entrants such as Apple have tried to keep a better defined line between the open source foundations of their programs and their proprietary extensions.

Other individual parts of Berkeley Unix continued to flourish as parts of the free software ecosystem. For example BIND, the Berkeley Internet Name Daemon, continued to be maintained by its original author, Paul Vixie[5] under the auspices of the Internet Software Consortium. Despite many commercial implementations, the open source version of BIND continues to be the definitive version that runs the Internet's Domain Name System (DNS), the single most mission-critical piece of software in the Internet infrastructure. Sendmail, another piece of Berkeley Unix, continues to be maintained by its creator, Eric Allman, who founded a company in 1998 to commercialize the software. He adopted a hybrid proprietary/open source strategy, completely consistent with the licenses, in which some new features of interest to commercial clients are released in proprietary software, while the open source version is also still maintained.

In short, research-style licenses, like the BSD and MIT Licenses, are ideal for situations in which you want wide deployment of your ideas and do not care whether this results in open source software or proprietary software. Because of their openness to commercial use, the programs they license can be, by many metrics, more influential. Red Hat maintains a Linux business that makes approximately $90 million in annual revenues, while Sun Microsystems has revenues of approximately $18 billion. There are literally billions of dollars of economic activity associated just with the Internet software stack originally released under the Berkeley License.

Nonetheless, the very success of the commercial developments premised on programs distributed under these licenses could be said to undermine the purpose of open source licensing. The argument could be made, for example, that the widespread adoption of commercial versions of such programs discourages open source development and encourages the creation of code closed off to the open source community by proprietary licenses. It could be regarded as a failure that the highly sophisticated Solaris software was developed as proprietary software,[6] that Microsoft was able to build a version of MIT's Kerberos security software that contains proprietary extensions for communicating with Microsoft servers, or that Microsoft was able to build so easily on the Internet infrastructure software.

It cannot be said, however, that such a result is inconsistent with the text and the intent of these licenses or that such types of commercial uses were not foreseen by their drafters. The original BSD and X Window System developers intended their software to be used in this way. Some of these developers even built their own companies based on the open source software that they had originally written. Bill Joy was one of the founders of Sun Microsystems; Eric Allman was able to found Sendmail, Inc.

The one well-known case in which the software authors were unhappy with their choice was the licensing of the MIT Kerberos security program. As Microsoft appeared to embrace and extend Kerberos, the authors wished they had used a more stringent license like the GPL. Of course, in that case, Microsoft would have chosen another basis for their security software, and Kerberos would have been less widely used. Nonetheless, the authors may have reasonably felt that a more restrictive license might have better protected the development of the software that they had anticipated.

Moreover, at least for certain types of programs, the nature of the function performed by the software makes additional license restrictions unnecessary to maintain an open development model. The Apache license provides one such example. While there have been several proprietary commercializations of Apache (such as the SSL-enabled Stronghold), the free version of Apache has retained its dominant market share as the result of two dynamics:

  1. Strong branding. The Apache License's requirement that derived works cannot use the Apache name gives a significant degree of protection.

  2. Standards-compliance. Because Apache is communications-oriented software, its need to adhere to standards such as the HTTP protocol prevents proprietary extensions. Of course, this protection remains only as long as Apache or other standards-compliant web servers retain dominant market share. Were Apache to lose its dominant market share, its protocols would no longer control, and this advantage would disappear.

These licenses, like all open source and free software licenses, permit forking and the subsequent fragmentation of projects. The multiple, and mutually incompatible, versions of BSD (FreeBSD, NetBSD, OpenBSD) provide one such example. However, this is less a result of the dynamic of the license itself than it is of the complex social dynamic involved in large software projects. The original BSD project leaders moved on to other activities, and the software was taken up by new people with different goals. This dynamic is discussed in more detail in Chapter 7.



[5] The specifications for this program were written by Paul Mockapetris.

[6] Sun has recently announced that it will release Solaris under an open source license, a major victory for open source.

Get Understanding Open Source and Free Software Licensing now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.