[ This is a version of an essay that I originally published on Conservancy's blog ].
Eleven days ago, Conservancy announced Kallithea. Kallithea is a GPLv3'd system for hosting and managing Mercurial and Git repositories on one's own servers. As Conservancy mentioned in its announcement, Kallithea is indeed based on code released under GPLv3 by RhodeCode GmbH. Below, I describe why I was willing to participate in helping Conservancy become a non-profit home to an obvious fork (as this is the first time Conservancy ever welcomed a fork as a member project).
The primary impetus for Kallithea is that more recent versions of RhodeCode GmbH's codebase contain a very unorthodox and ambiguous license statement, which states:
(1) The Python code and integrated HTML are licensed under the GPLv3 license as is RhodeCode itself.
(2) All other parts of the RhodeCode including, but not limited to the CSS code, images, and design are licensed according to the license purchased.
Simply put, this licensing scheme is — either (a) a GPL violation, (b) an unclear license permission statement under the GPL which leaves the redistributor feeling unclear about their rights, or (c) both.
When members of the Mercurial community first brought this license to my attention about ten months ago, my first focus was to form a formal opinion regarding (a). Of course, I did form such an opinion, and you can probably guess what that is. However, I realized a few weeks later that this analysis really didn't matter in this case; the situation called for a more innovative solution.
Indeed, I recalled at that time the disputes between AT&T and University of California at Berkeley over BSD. In that case, while nearly all of the BSD code was adjudicated as freely licensed, the dispute itself was painful for the BSD community. BSD's development slowed nearly to a standstill for years while the legal disagreement was resolved. Court action — even if you're in the right — isn't always the fastest nor best way to push forward an important Free Software project.
In the case of RhodeCode's releases, there was an obvious and more productive solution. Namely, the 1.7.2 release of RhodeCode's codebase, written primarily by Marcin Kuzminski was fully released under GPLv3-only, and provided an excellent starting point to begin a GPLv3'd fork. Furthermore, some of the improved code in the 2.2.5 era of RhodeCode's codebase were explicitly licensed under GPLv3 by RhodeCode GmbH itself. Finally, many volunteers produced patches for all versions of RhodeCode's codebase and released those patches under GPLv3, too. Thus, there was already a burgeoning GPLv3-friendly community yearning to begin.
My primary contribution, therefore, was to lead the process of vetting and verifying a completely indisputable GPLv3'd version of the codebase. This was extensive and time consuming work; I personally spent over 100 hours to reach this point, and I suspect many Kallithea volunteers have already spent that much and more. Ironically, the most complex part of the work so far was verifying and organizing the licensing situation regarding third-party Javascript (released under a myriad of various licenses). You can see the details of that work by reading the revision history of Kallithea (or, you can read an overview in Kallithea's LICENSE file).
Like with any Free Software codebase fork, acrimony and disagreement led to Kallithea's creation. However, as the person who made most of the early changesets for Kallithea, I want to thank RhodeCode GmbH for explicitly releasing some of their work under GPLv3. Even as I hereby reiterate publicly my previously private request that RhodeCode GmbH correct the parts of their licensing scheme that are (at best) problematic, and (at worst) GPL-violating, I also point out this simple fact to those who have been heavily criticizing and admonishing RhodeCode GmbH: the situation could be much worse! RhodeCode could have simply never released any of their code under the GPLv3 in the first place. After all, there are many well-known code hosting sites that refuse to release any of their code (or release only a pittance of small components). By contrast, the GPLv3'd RhodeCode software was nearly a working system that helped bootstrap the Kallithea community. I'm grateful for that, and I welcome RhodeCode developers to contribute to Kallithea under GPLv3. I note, of course, that RhodeCode developers sadly can't incorporate any of our improvements in their codebase, due to their problematic license. However, I extend again my offer (also made privately last year) to work with RhodeCode GmbH to correct its licensing problems.
