I have decided that I think that it is necessary at this point to put my thoughts together regarding the GPL, and when licence violations can gain the force of copyright violation. I am not a lawyer, but this has come out of watching a number of cases, discussing the issue with a number of lawyers, and trying to understand all sides.
When a GPL violation case comes up, folks generally are quick to argue that it is definitely copyright infringement. Stallman has even argued that nVidia's drivers infringe on Linus's copyrights. While I think that a subset of GPL violations do rise to the level of copyright infringement, I think these cases are somewhat overstated.
The GPL, despite what Stallman says, is a contract in which both parties agree to abide by certain behaviors in joint interest. The contract is an adherence contract similar in force to a web-site's terms of service (where use of the good or service requires adhering to the contract) and the consideration found is in the requirement of equal access to pulically distributed code. The GPL is much more like a contract than are more permissive licenses, like the BSD license, because the consideration factor is quite a bit greater. For example, while the BSD license might be argued not to include consideration since the only requirements are those required minimally by copyright law (not stripping copyright headers) and (when distributed in source form), not making false claims about warranties, the GPL actually requires the licensee to share something further with the idea that it will be available to the original developer. "I will share if you will share" is consideration while "I will share, but don't say I am giving a warranty when I am not" might be argued not to be. Similarly, the 4-clause BSD license (with the advertising clause) isclearly a contract, while the two-clause BSD license might not be.
At the same time, it seems reasonable to argue that a contract violation regarding copyright terms could become a copyright violation if the behavior is sufficiently outside the scope of the license. For example, if I grant someone a license to publish five copies of my book for a flat fee of $20, and they publish 5000 copies of the book, that would seem to be copyright violation, not a mere contractual issue. At the same time, I don't think it is copyright infringement if there is a reasonable argument to be made that the contract allows the use, or if the difference is small enough as to represent an issue resolvable through contract dispute (you print 6 copies instead of five by accident, that should be a contract matter). And certainly a mere reasonable disagreement as to the terms of a contract should not subject the loser in the case to copyright infringement sanctions.
Lawyers in contentious cases tend to find as many areas to allege misbehavior and as many grounds for relief as possible. Consequently, one can expect that any case of stepping outside of the perceived boundaries of a license will be labelled as copyright infringement because of the chance that the court will find for the plaintiff on this matter. It is thus understandable that lawyers will raise this issue in minor contractual disputes for leverage.
One of the most interesting cases which provides a parallel currently is United States v. Lori Drew. In this case, the US Attorney involved is seeking criminal sanctions over terms of service violations on MySpace's web site. Lori has been convicted of three misdemeanor counts of computer hacking for violating MySpace's terms of service (and creating a fake profile). Currently the court is considering throwing out those convictions in a directed verdict motion. If not, the next step is the 9th Circuit Court of Appeals. The judge is obviously having a hard time with the ruling since sentencing has been delayed for a total of seven months while he considers the motion to acquit. The key element here from many who support dismissing is that web site terms of service violations should simply not be prosecutable as crimes. Many of us feel that turning any term of service violation into a crime is dangerous to our system of law, and the same occurs with any other adherence contract. To hold the GPL to a different standard than MySpace's terms of service just because we like the license is hypocritical and similarly dangerous.
What I would propose in these cases is the concept of a penumbra around contracts, where violations are merely contractual issues. The penumbra would be defined both in terms of severity of the violation and vagueness in the contract. Any reasonable argument that the behavior was allowed in the contract would be sufficient to place the behavior under the penumbra where contractual violations could not lead to further legal or statutory challenges, as would the argument that the violation was not particularly egregious.
Back to my book analogy... Suppose in addition to limiting the number, I also require the book to be distributed on media suitable for being input directly into a computer or an offer valid for three years to provie such. Suppose the publisher does this by typesetting the book in the OCR-B font, and arguing that this is suitable for optical scanning and therefore they have met their terms under the contract. I take them to court. I don't think the court should entertain the notion that there are copyright violations in this case because there is a reasonable argument to be made that printing the book in a medium designed for both humans and computers is allowed by the contract. If I ultimately prevail, it should be on intent of the contract, and it should be a contractual matter.
So the next issue becomes the question of whether the GPL can regulate bridges (via linking) between a GPL application and a closed source application. Stallman says such bridges (such as the LGPL components of the nVidia drivers) are not in line with the license. He raises arguments which seem to be similar in nature to the arguments raised by The SCO Group in their suit against IBM. The major questions are:
1) Does linking NECESSARILY imply derivation?
2) Is derivation contageous? I.e. if A is derivative of B, and B is derived of C, can we say that A is derivative of C without further evidence?
Regarding the nVidia Driver issue, the typical understanding is that nVidia has ported the core logic of their windows drivers into a module which is independant of the Linux API itself. nVidia then provides a Linux driver, under the LGPL, which links the Linux kernel and the closed source module together and handles how the Linux kernel interacts with the closed-source module. Assuming this is the case, it would seem that nVidia has actually fulfilled their rights under the GPL v2. The reasons are elucidated by a reading of various rulings in SCO v. IBM.
In SCO, the court ruled that derivation was not contageous, and that one must show a continuity of the expressive elements in order to find derivation. In short, if A is derivative of B, and B is derivative of C, in order to say that A is derivative of C, one must show actual structures in A that are derivative of structures in C. It seems unlikely, given the standard understanding of this case, that the nVidia drivers in fact are derivative of the Linux kernel in this way, so they are not bound by the GPL. Similarly, under the GPL v3, it seems to my mind that one could easily create such a bridge without running amok because one can add additional permissions to specific modules (or even license modules under more permissive licenses like the BSD license).
The second issue, however, is the question of whether linking is decisive in the derivation discussion is an interesting one and has been dealt with substantially in other papers (see previous journal entries for citations). The general attitude seems to be that linking does not by itself imply derivation though it could lend some weight to the idea, particularly where object-oriented techniques like inheritance are used. However, a lack of linking does not mean that a work is not derivative either, particularly in more expressive content such as game displays (altering a game display, however done, might well be seen as creating a derivative work).
However, even if such a view were to be frowned on by the courts, I would hope they would see reasonable arguments to the contrary as requiring damages based solely on contractual violations rather than copyright infringement. Either way, I think Stallman is wrong and is advancing some dangerous arguments which we are rightly wary of in different cases. The key issues for me are: 1) do most lawyers I know accept those arguments, and 2) would we feel differently if those arguments were being advanced against Free/Open Source Software?