After 18 months of widespread consultation with community and corporate interests, the third versions of the GNU General Public License (GPL) and GNU Lesser General Public License (LGPL) were released one year ago on 29 June 2007. In November, they were joined by the GNU Affero General Public License (AGPL). Looking back at these licenses today, observers of free and open source software (FOSS) judge them a modest success, and credit them with continuing to educate people about free software.
Figures on adoption of the GPL family vary somewhat, depending on whom you talk to. According to Doug Levin, president and CEO of Black Duck Software, a company that tracks open source code, 2,476 projects are now using the third version of the GPL, 358 are using the LGPL, and 72 the AGPL. Theresa Bui, vice president of marketing for Palamida, a firm with a similar mission, gives roughly similar figures of 2,271 for the GPL, 261 for the LGPL, and 100 for the AGPL. The difference in the figures is probably explained by the fact that Black Duck's figures include planned transitions to the licenses, while Palamida's does not. Palamida's higher figures for the AGPL are probably due to Palamida's recent interest in the license, as evidenced by its blog.
In addition, Palamida tracks figures for projects that use a version of the GPL that contain a clause allowing contributors to use version 2.1 or later of the license (which includes, of course, version 3.0). These figures are 6,467 for the GPL and 372 for the LGPL. Since the AGPL was not in widespread use until the third version, it is not included in these figures.
Despite the difference in figures, both Black Duck and Palamida agree on the overall trends. Both agree that, taking all versions and variants of the GPL as a single unit, they account for about 70% of all the free licenses in use. Moreover, both agree that the adoption rate for the third versions of the GPL and LGPL remain steady at about 220 per month, and that this growth does not seem to have come at the expense of the second versions. Both expect these trends to continue with little change for the rest of 2008. Neither expects the third versions to overtake the second ones any time soon.
The major difference in adoption rates comes with the AGPL. Levin is cautious about talking about trends with AGPL, pointing out that it has only been out for six months. He says only that "there has been some pick up on the adoption of the AGPL." By contrast, Bui says that AGPL adoption "is really starting to take off." She predicts another 50 projects will be using the license by August, and describes this estimate as "incredibly conservative." She goes on to say that the AGPL is causing far more excitement than the latest versions of the GPL or LGPL, most likely because it received less attention in the consulting process that produced the third versions, and therefore came as more of a surprise.
By contrast, Bui says that the slow but steady adoption rates of the GPL and LGPL third versions was predictable. Instead of rushing to switch to the newest licenses, she says, "Software projects have been just chugging along in their old, familiar lifecycles, and, when new versions were ready, they converted to GPLv3."
One noticeable trend in adoption of the third versions is that projects and companies that opposed them or were neutral during the consultation process do not seem to have changed their minds. "It seems that people who had significant qualms about it a year ago haven't moved from them," says Bui. As a result, while some prominent companies and projects have moved to the third versions, including SugarCRM, Sun Microsystems, and OpenOffice.org, others, equally influential, have not, including the Linux kernel and MySQL.
The outspoken resistance by Linux kernel developers that was so highlighted during the consultation process may be slowing the adoption rate. "The fact that the kernel has not adopted GPLv3 is a significant part of any group's decision not to adopt," says Bui.
Still, both Bui and Levin see the third generation of the GPL family as being steady, if unspectacularly successful. "We see no signs of trends increasing or decreasing for the rest of the year," says Bui. As for the second generation, she predicts that "it will not, in the next two years, lose its position as the number one license."
Similarly, Levin says, "I believe that the GPL will have an average increase of about 10% per month over the next year. What that means is that there will be some 6,000 projects using the licenses a year from now. It will continue to grow and continue to be a factor in the mix of licenses, but the other licenses will continue to be used."
Another measure of success
However, statistics are only one way to measure the licenses' success, as Peter Brown, executive director of the Free Software Foundation, emphasizes. While Brown says he is "happy with the rate of adoption," he suggests that the more important proof of the licenses' success is their ability to educate free software users and "prevent free software from being made proprietary. In that sense, we are very happy about it."
Nor is Brown unduly disturbed that the second versions remain the dominant ones, even though they offer less protection than the third ones. "It's not as if there's been a move away from the GPL," he notes. "It's the same GPL, in a sense."
Instead, Brown suggests that the consultation process and the third versions themselves "have helped the conversation about free software progress." The fact that large sections of the FOSS community were consulted about the wording of the new licenses helped both to educate people about the issues surrounding them, and to give them a stake in the process.
In addition, the Free Software Foundation has made considerable efforts to educate users about the licenses, such as its page about the advantages of the third version.
These suggestions are supported by Bui's observations at client sites of people using their knowledge of the GPL licenses to help plan product lifecycles. "That's due to the good work of the FSF," she says. "They have done a really good job of educating people about the GPL and raising their awareness."
Brown also suggests that the lack of legal challenges to the third versions may be a measure of their success, although he admits that they may be too recent to be contested, and that the second versions have not received many challenges, either. But he credits the consultation process with improving the legal language of the licenses, as well as making clear from the number of people involved in the consultation just how large a community any potential violators would be taking on.
Still another factor may be that the technology field has caught up with the licenses. When the licenses were being drafted in 2006-07, lockdown technologies were just starting to appear, so the language to prevent them was seen as unnecessarily radical. Now, with examples like the iPhone, such provisions are starting to seem far-sighted.
"The main thing is, we're satisfied that the changes that we've made seem to have done their job," Brown says.
Future measures of success
A year after third versions began to be released, an absolute judgment of their success is still impossible. No comparison can be made with the success of the second versions after their release in 1991, because the free software community and the pressures upon it have changed so drastically since then.
For another thing, the licenses remain a work in progress. One of the changes in the third version is greater allowance for exceptions -- special clauses that make small modifications to the basic GPL license. Brett Smith, FSF compliance engineer, is still working with the community to modernize and improve exceptions for the third generation. "GPLv3 came out on a Friday, and I came in Monday and started working on the exceptions," Smith says. A year later, they still occupy much of his time, but their effectiveness will be a major factor in how well the licenses themselves function.
It will take several years to see exactly how good a job the third versions do of "futureproofing free software," as Brown puts it. He suspects a major test will be the ability of the AGPL to create and safeguard free network appliances. Nor does he rule out the eventual need for a fourth generation of licenses in response to the discoveries of new loopholes.
For now, the only measure of success is that the third generation of the GPL family seems to be holding its own. It has not replaced the second generation, as many might have assumed it would when the process of creating it first began in 2006. Still, if counted separately from the second generation, in a year it has become the seventh most popular free license, according to Black Duck's statistics.
As Brown points out, "GPLv3 is probably the most adopted license in the last year. And it seems to achieve its aims thus far." For any other license, this performance would be a runaway success. It is only by comparison with the dominance of the second version of the GPL licenses that the success of the third versions might seem disappointing.
But the ultimate success of the third generation GPL, and the question of whether it will replace the second generation or continue to coexist with it, remain almost as uncertain today as they were a year ago.