“The gist is this: Under the terms of their initial GPL, Holovaty and the other developers can do whatever they want with their code (after fulfilling the Knight requirement) because they hold copyright. There’s additional nuance brought about by the fact that Everyblock is a web service and not and piece of ‘software’ per se.” – Chris Anderson, Journalism PhD, “The Nuances of the Everyblock Sale to MSNBC”
There are two pieces in there that need to be taken apart: the subject of copyright, and the nature of EveryBlock. The second piece is, by comparison, simple. What Anderson presumably means is not that EveryBlock is not a piece of software, but rather that it’s a piece of software that will not be distributed in the traditional sense and will not, therefore, trigger the reciprocal provisions of the GPL. The provisions, more specifically, that insist that changes and updates to a GPL’d asset be made available under precisely the same license.
While it’s clear that Anderson, like many non geeks out there, is unfamiliar with the free and open source worlds and their respective licensing concepts, he’s grasped the basics of the EveryBlock situation correctly. Short of the seriously underadopted AGPL, there are no licenses that would have prevent the quote unquote closing of the EveryBlock codebase because it’s about copyright – Anderson’s first point – not the license.
Which, upon reflection, is probably all the response I need to Brian Aker’s presumably tongue in cheek followup question to my “Does the GPL Matter?” bit. As an author of MySQL and Drizzle – not to mention the recipient of this year’s Best Open Source Database Hacker award at OSCON – there are few who know the answer to that question better, but it’s worth exploring for the rest of us. Because the implications of copyright ownership, as Anderson’s lament on behalf of the Knight Foundation perfectly illustrates, can be both surprising and frustrating for those unfamiliar with where licensing ends and copyright begins.
What anyone familiar with the history of MySQL already knew, and many journalists like Anderson are becoming aware of, is the reality the license is but one of the legal mechanisms restricting the unfettered usage of any given codebase. Copyright also has its role to play, as do trademarks. And while most of the media attention is paid to the licenses, and the differences between them philosophically, there are a variety of circumstances in which it’s the trademark or the copyright that’s at issue. But whether it’s an issue depends, as we’ll see, on the governance structure.
In open source, there are two basic models for copyright ownership: single entity and community. MySQL is the canonical example of the former, Linux, the latter. Single entity models own the entirety of the copyright for the asset they maintain. Either they’ve written the entirety of the code, or they’ve acquired – either through commercial licensing or some form of a Joint Copyright Agreement (JCA) – the rights to it. This means, essentially, that they are free to operate independently of whatever license they choose to apply to the asset. In the case of MySQL, they retain the ability to ignore the GPL they’ve applied to the project and exclusively relicense the code for commercial gain. If they wished to continue development of the asset under a closed source, proprietary license, they would be entitled to do so. They couldn’t revoke the rights originally granted to the codebase, of course – those remain in perpetuity. But they could proceed by developing only closed source future versions of the product, because they own the copyright.
Linux, by contrast, is such a maze of copyrights that there have been threads on whether relicensing would be even theoretically possible given the fact that some of the original copyright holders were deceased. Unlike MySQL – or, as Anderson points out, EveryBlock – the copyrights to the code that makes up Linux is held in commons, such that no single entity – not even the majority holder, Linus – can make exclusive use of it. SuSE, for example, can’t introduce its own proprietary extensions to differentiate itself from Red Hat. Oracle can’t introduce a compatibility layer that only works with its distribution of the kernel. And so on. The rights enjoyed by one are the rights enjoyed by all, because the copyright doesn’t belong to any single party.
So the answer to the question of whether copyright ownership matters is simple: it is self-evident to me that it does. The next question, however, is one I’m less prepared to answer: are single entity copyright ownership models viable going forward?
For while MySQL has long been the standards bearer for this approach – and the economic model that accompanies it – it’s not apparent that that approach will be indefinitely sustainable for the project. Which in turn begs the question: what would the implications be if MySQL, of all projects, were forced to abandon the dual-licensing model it had long championed? Would dual-licensing as we know it be practicable?
It’s difficult to say, but I think it’s certainly possible that we could see dual-licensing/single-entity copyright ownership relegated to a bootstrapping strategy, one as designed and structured for obsolescence as our baby teeth. Because while projects the size of EveryBlock are unlikely to face serious competition in terms of understanding and have the ability to extend the asset, projects the size and importance of MySQL will, and are.
Put differently, over the lifetime of a successful open source project, it’s likely that single entity copyright ownership will gradually cease being an asset and become a liability. Adoptees of this model, then, would do well to understand and plan for this shift, lest they become irretrievably dependent on an increasingly problematic development model.
Thus, for better and for worse, copyright ownership matters. A lot.
Disclosure: MySQL, and its parent company Sun, are RedMonk customers, as is Red Hat. SuSE, and its parent company Novell, are not; neither is Oracle.