Don Rosenberg's review in LWM (Vol. 3, issue 4) of Larry Rosen's book, Open Source Licensing, did double-duty as a platform for FUD about the GNU GPL.

The GNU General Public License (GNU GPL for short) was not the first free software license, but was the first to embody the concept of "copyleft": the requirement that all modified and extended versions of the program be free under the same license. The GPL was designed to use copyright law to defend to the utmost all users' freedom to copy and change software; since copyright law says one must get permission to use a work, either in derivative works or combined works, the GPL requirements apply to both cases. Any change in a program can be made by changing its existing modules, or by adding a new one; the choice is arbitrary. Therefore, an effective copyleft license must cover both ways of structuring the changes. The GPL does so.

In the 16 years since version 1 of the GNU GPL, it has become far and away the most popular free software license. (Those who advocate open source may note it is also far and away the most popular open source license.) We have another license, the GNU Lesser GPL, that permits linking with non-free modules. It's used much less often, and that's as it should be. Any of these projects' developers could have chosen the LGPL, but most of them did not.

Developers select the GNU GPL for various reasons. Free software activists wish to create the biggest possible pressure for others to make their work free software. Some software businesses want to make users pay to link their free software products with non-free software.

Many developers simply don't want to have to compete against non-free extended versions of their own programs.

Generally speaking, the GNU GPL serves our community well. For instance, GCC includes C++ support only because of this. Since the C++ front end had to be linked with the GCC back end, its developer could not make it proprietary.

Generally speaking, the users of the GPL are happy with it, but Don Rosenberg is not happy. His review suggests that the Free Software Foundation "solve the GPL's problems" by converting it to something like the LGPL - in other words, by gutting its most important strength.

Such a change would mean abandoning our goal: that computer users should have the freedom that free software gives them. It would also betray the thousands of developers who have chosen the GNU GPL, not the LGPL, because they want to defend the freedom of all versions of their software. Whatever problem there may be, this solution is worse.

What are the problems that we could supposedly solve by destroying our work? Some have been fabricated, others exaggerated.

Rosenberg, like Rosen, claims that the GPL's requirements on combined works (i.e., extended versions) cannot be enforced. Our legal advice says they can be; and the hundreds of violators who chose to correct their violations rather than challenge us in court seem to agree. The reason is simple: as Rosenberg writes, "the collector is the author [of the collection] and can license and distribute it, but only if he has a distribution license from the authors of the constituent pieces." This license, for a GPL-covered piece, is given only when the combination will be released under the GNU GPL or consists of mere aggregation of separate and unrelated programs.

Rosenberg claims the GPL is "ambiguous" about linking GPL-covered code with non-free code, and that judges would reject it out of hand; but when MySQL went to court against NuSphere, in just such a case, it prevailed.

Rosenberg claims that the GPL is a failure because businesses reject GPL-covered software. If they really did, that would be unfortunate but not disastrous; success in winning freedom is not measured by the good opinion of business. However, this claim appears to be mistaken.

The fact is that GPL-covered programs such as GCC, Linux, and MySQL are quite popular in business. The GNU/Linux operating system, which contains those programs together with thousands of other GPL-covered components, is becoming more popular every year.

Rosenberg cites four of Larry Rosen's criticisms of the GPL; all four are erroneous as stated, though it is true that compatibility with additional free software licenses would be desirable. In particular, the GPL does not try to "convert collective [he means combined] works into derivative works"; it simply applies to both. Object-oriented programming does not cause a problem for the GPL, and Rosenberg is mistaken in claiming we have tried to solve one.

I have mentioned only a handful of the numerous errors. Any time Rosenberg says the FSF did something, or believes something, or aims for something, I suggest checking www.gnu.org, or asking the FSF directly, before believing the claim.

Anyone can be mistaken, but the review shows a pattern of spin as well. It's normal to prepare a draft text you are happy with before asking others to comment on it, but he finds this sinister when we do so. The fact that we negotiate with violators rather than rushing into court is sound practice, but he finds in it proof of hypocrisy.

Most telling of all, he says we should gut the GPL to be more "modern" - the argument from latest fashions, a common form of irrational persuasion. This forms a pattern that suggests a wish to find fault with the GPL, which could be responsible for some of the errors.

GPL version 2 is not perfect, and we are working on making improvements in certain details. For instance, we hope to achieve compatibility between GPL version 3 and a future version of the Apache license. I plan to devote the last quarter of this year to finishing up a draft ready to show. Those who wish to see the GPL gutted will surely be disappointed, but anyone who is happy with GPL version 2 will probably like the changes.