GPLv3 adoption on track, experts say


Author: Bruce Byfield

How is the third version of the GNU General Public License (GPLv3) being received four months after its official release? Not well, if you believe the Evans Data survey released on September 25. However, those who concern themselves with licensing issues at the Free Software Foundation (FSF) and at Palamida, a company that advises customers on issues that surround free and open software (FOSS), paint a different picture. According to these FOSS experts, adoption of GPLv3 is going as expected, and, while reasons for caution exist, the new version is likely to replace GPLv2 some time in the next few years. As for the Evans Data survey, they suggest that the information released overgeneralizes a complex situation.

From its survey of 380 developers, Evans Data concluded that “only” 6% of developers working on FOSS have adopted GPLv3. Even more seriously, it claimed that two-thirds of those surveyed did not plan to adopt GPLv3 in the next year, and 43% would never switch. The company said that “almost twice as many would be less likely to join a project that implements GPLv3 than would be more likely to join,” although without giving exact figures in the announcement of the results. (The full report is available only to paying customers.) The implicit claim in the announcement seems to be that GPLv3 is a failure.

However, to those involved in FOSS, this claim seems premature. Both Theresa Bui, Palamida’s vice president of marketing, and FSF Executive Director Peter Brown agree that many analysts and community members have unrealistic expectations about the rate at which conversion to GPLv3 will occur.

Referring to the day after GPLv3 was released, Bui says, “Some people really thought there would be some giant avalanche of GPLv3 adoption on June 29. But those people are unfamiliar with how software is developed. Trying to do regression testing and cutting a whole new version of your software is not really among the top priorities of most software projects.” She suggests that many FOSS projects will switch to GPLv3 only with their next software release.

Similarly, Brown points out that talking about FOSS developers as a whole means little in this context. “That kind of survey only makes sense if you’re talking to people who currently support GPLv2,” he says. After all, while GPLv2 is one of the most popular FOSS licenses — accounting for up to three-quarters of all FOSS projects according to the most commonly used figure — that still leaves a large proportion of projects that use other licenses, such as the Apache or BSD licenses, for whom the question of using GPLv3 is a non-starter.

“I think you’ll find that most projects haven’t ruled out GPLv3 yet,” Brown says. “They simply haven’t had an opportunity to look at the reasons why they would move. So it’s a strange angle to come at the license from.”

Moreover, the statistics from Evans Data seem to contradict Palamida’s own record of GPLv3 conversions. When this article was written, Palamida’s records show 836 projects have converted to GPLv3, with another 82 moving to the Lesser GPLv3. These figures imply a slow but steady conversion rate.

Palamida also mentions 5,738 projects licensed under GPLv2 or later, which gives distributors the right to decide which license they choose. While these projects may not support GPLv3, they may also have no incentive to convert, since their license terms already allow the possibility of distributors using GPLv3.

The truth is, four months is far too early to suggest that GPLv3 is faltering. If nothing else, no comparison exists from which any conclusions can be drawn. “If this was a new release of Firefox or something like that,” Brown says, “you could go by historical trends. You could see if the latest release was as popular as the last one.”

The trouble is, while the GPL was revised before, that was in 1991, when FOSS was a fringe idea and the community was smaller and more unified.

“This is really the first time anything like this has happened in the free software community, even when Apache upgraded its license,” says Brett Smith, the FSF compliance engineer who regularly advises projects and companies on licensing. “The scale is so completely different that you can’t really make any kind of comparison. I personally didn’t have any sort of expectations going into this process because it really is unprecedented. But, despite that, I definitely feel that things are going well.”

Bui is more cautious, but generally agrees. “The adoption rate is clipping along at the normal rate that software is usually developed,” she suggests.

Concerns and delays

For all their wait and see attitude, FOSS experts also see some concrete issues that may be delaying GPLv3 adoption. Since last spring, Bui says, Palamida has been hearing concerns about three new additions in GPLv3: The language concerning lockdown technologies, the new provisions for patent protection for all users, and the clause designed to forestall redistribution deals like the one that Microsoft signed with Novell last November.

However, unlike Evans Data’s president and CEO John Andrews, FOSS experts do not believe that these additions are causing a problem simply because they impose “restrictions on what you can do with programs implemented under this license.” After all, much the same can be said of GPLv2 or, in fact, any license.

Instead, the issue is partly that these provisions are new, which is why, having released GPLv3, the FSF has recently been focusing on educating people about the license. To this end, Smith has already held one IRC discussion on GPLv3 issues, and says that more are planned. He is also considering conference calls and seminars aimed at both community projects and corporations.

According to Bui, the new license also coincides with a growing crisis in business attitudes towards FOSS. She says adoption of all free software “has outstripped the old traditional ways that procuring software has happened.” In other words, instead of being vetted for liability and security, the way that proprietary software has been, FOSS has been slipping into business largely unannounced, both because it is free and because the workplace is more decentralized than in the past. But, in recent years, she says, businesses “have started to realize the amount of undocumented code that has come into their networks and behind their firewalls. That makes them very nervous.” With this concern, the adoption of any new license, not just GPLv3, is now being treated more cautiously than in the past.

Moreover, both the FOSS community and business may be nonplussed by the 18 months of debate that accompanied the writing of the GPLv3. Apart from a general tendency of the community to be more open to conversion than business is, none of the FOSS experts otherwise saw much of a difference between the two groups in their overall willingness to consider using the GPLv3.

Jeremy Allison explains the Samba project’s quick adoption of GPLv3 by saying that “Members of the Samba team participated in the drafting of the GPLv3, and we’re very happy with the way it ended up.” However, similar causal relations are hard to find.

Still, the writing of the GPLv3 and the debate around it does seem to have made the community and corporations alike aware of the divisions among those involved with FOSS. The debate, Bui says, showed that those involved in free software today now range from political and philosophical idealists to those whose main interest is openly capitalistic.

Both Bui and Brown point in particular to Linus Torvalds’ outspoken defence of his decision to keep the Linux kernel under GPLv2. “Linus has become the voice of the anti-GPLv3 sentiment,” Bui says, “and Richard Stallman the voice for the pro-GPLv3 sentiment. And of course people will pick that up, because those are two very polarized opinions. Yet those two men and those communities and the code they represent are the quintessential essence of open source and the Linux platform.”

With the possibility of project forks because of license disagreements, many FOSS users seem to have become cautious. Nor is the uneasiness helped, Bui suggests, by the blunt, often rude rhetoric in which these disagreements were conducted — although she adds that those involved in software production are probably less affected by the language of the debate than those who simply use FOSS. simply because they know enough to discount much of it. Community members are likely to be experienced enough to discount much of the rhetoric, but to outsiders, this rhetoric can make disagreements appear more serious than they are.

For a handful of FOSS contributors, adoption of GPLv3 has not occurred largely because the details are still being worked out. Even in the FSF’s GNU project, some teams have not switched to GPLv3 because they are still working out the language for the exceptions they require in the license. For example, Smith notes that the GNU Compiler Collection requires an exception written into its license so that the code it leaves in compiled programs does not make every program that uses it subject to the GPLv3.

However, even where none of these concerns exist, Brown suggests that GPLv3 adoption was never likely to be immediate. All the potential issues need to be discussed separately in each project, and, although the FSF is the owner of the license, it has no power to insist on its use. “We’re not in control of other people’s licensing,” Brown says. “Our role isn’t to push use of the license — it’s to advocate, and the only way we can do that is by talking to individual projects, which takes times.”

Predictions and measurements of success

For all her awareness of the problems, Bui suggests that the GPLv3 will probably replace GPLv2 in “the next 12 to 18 months.” For Bui, the key to GPLv3’s success is the core GNU utilities that form a part of countless distributions and other software. “My prediction,” she says, “is that, if core non-branded. heavily reused projects like any of the GNU projects continue to be adopted and reused as they have been, then GPLv3 will become the dominant license merely because of the fact it’s going to be embedded in so many open source projects that they, too, will be GPLv3. GPLv3 will become dominant.” The only factor that might disrupt this process is the development of equivalent utilities under another license.

Similarly, Brown says that GPLv3 adoption is “a process that will take years to occur. These considerations will take time to mature and for people to understand what’s going on behind the license.” For Brown the real question is what situation free software may face in the future. “If people are in an environment where they feel the kind of freedoms we’re trying to defend are not under threat — where the environment has become safer for free software — then maybe they won’t be rushing to change to GPLv3. But it really comes down to time for projects to consider the license, the environment that people find themselves in, and the educational aspects.”

In the end, for the FSF, the measure of the GPLv3’s success will be, not how widely it used, but whether it can protect software freedom. “What does success mean?” Brown asks. “I think it’s important not to set some arbitrary target about what success means. It’s not a numbers game. It’s about how do you get people to value the aims of GPLv3. Because if you don’t value the freedoms of the GPL, then the question of GPLing isn’t about freedom. It’s a question of technical considerations.”


  • Legal
  • Free Software