On the Decline of the GPL

Share via Twitter Share via Facebook Share via Linkedin Share via Reddit

Guess which open source license is more popular than the MIT, Artistic, BSD, Apache, MPL and EPL put together? Surprise: it’s the GPL. True, usage appears to be in steep decline. Since August of 2009, the GPL is down around 8%, according to data from Black Duck. Over that same span, usage of permissive licenses is up: MIT by 8%, Apache 2% and BSD 1%. But while developers may be increasing their usage of non-copyleft licenses, is this a problem?

With all due respect to Red Hat counsel Richard Fontana, for whom the waning usage of the GPL appears to be alarming (as an aside, I find it mildly ironic that the project that built those slides is itself permissively licensed), this seems to be little more than a normal market adjustment. The unnaturally dominant role copyleft licensing played for many years was, in my view, as much an artifact of the extraordinary visibility of projects like the Linux kernel and the MySQL database as project owners’ affection for the reciprocal protections offered by copyleft licensing. As such, it never appeared to be sustainable from this vantage point, which is why we predicted back in 2009 precisely what has occurred: gains from permissively styled licenses at the expense of reciprocal alternatives.

The GPL is an enormously important mechanism, as we’ve asserted since at least 2005. It simply could not expect to be the only mechanism, indefinitely. Licenses are tools, and should be selected and employed based on a desired outcome. As those desired outcomes have changed over time, it’s only logical that licensing patterns change to accomodate.

We have argued, both observationally and based on public market valuations, that the value of software as a differentiated asset is in decline. The evidence suggests that native web businesses assign a substantially lower value to written software than did their predecessors. Facebook, for example, originally wrote Cassandra to manage their Messages feature, subsequently releasing the code into an Apache project. When they rebuilt Messages, they chose Hbase – an Apache project originally created by a separate organization, Powerset – over their own Cassandra. GitHub’s Tom Preston-Werner, for his part, recommends open sourcing all but those features that represent “core business value.”

What both organizations have realized is that very little code, in practice, is competitively differentiating. Which makes open source a logical course of action, because the potential benefits of making the source code available are likely to substantially outweigh the costs. And as far as licensing is concerned, if the code is not a competitive advantage, it is likely not worth protecting. For those who view the code they produce as a generally fungible asset, the additional protections afforded by a reciprocal license may not only be unnecessary, but unwanted. In this scenario, permissive licenses are a perfect alternative.

Which should be ok. Open source licenses are, ultimately, different tools. Employing them towards different ends is nothing more than logic.