I've generally agreed with theme authors that just interoperating with GPL software via its API is not enough to require your work to also be GPL'd, but when your theme or plugin contains whole sections of modified GPL code you've pretty much lost the argument.
But, this still doesn't make it OK for Automattic to force all theme developers to move to GPL. As that guy rightly points out, by the same logic you should sue all closed source software on Linux that makes use of system calls.
GCC itself is GPL, but its libraries (along with glibc) have an exception that permits linking against non-GPL software.
That they consider it a derivative work is their -legal opinion-. Remember their license still doesn't supercede local laws and a court could invalidate Wordpress' legal opinion of what a derivative work is.
NOTE! This copyright does *not* cover user programs that use kernel
services by normal system calls - this is merely considered normal use
of the kernel, and does *not* fall under the heading of "derived work".
Then there is still a problem with the exception. It only applies to code owned by the kernel developers. If I were to come along and contribute, say, a driver to the kernel, I'd have to agree tot he exception and license my driver as GPLv2 + exception. No problem.
But suppose as part of my driver, I needed some code for some data structure. I find that someone has already wrote such code and put it in a GPLv2 program. I'm using GPLv2, so I take their code and put it in my driver.
But wait! The author of that data structure code has not agreed to GPLv2 + exception. He's just released under GPLv2. I don't own the copyright, so can't relicense it under GPLv2 + exception.
If the kernel developers accept my driver, then the author of that data structure code could take action about people running non-GPL applications on Linux.
How careful are the kernel developers and maintainers about this? Do they make sure that anyone who contributes code, even indirectly, has agreed to allow the exception?
This in general is an issue with any code that uses a GPL + exceptions license, if the exception is actually needed. (Many feel that the Linux exception is NOT actually needed)
Such a practice is used in other projects ... e.g. the GNU Classpath exception ... http://www.gnu.org/software/classpath/license.html
BTW ... a piece of software X is considered a derived work of some software Y, if X is dependent on Y for functioning and Y can't be replaced.
Linux may not need that exception for many user applications, but without it you're in gray area.
The notion of a work "derived" from another work is a notion from GPL, not from copyright law. The copyright notion most closely corresponding to what GPL seems to mean by derived work is "derivative work".
Whether or not X is a derivative work of Y does not depend on whether or not Y can be replaced. For instance, if I sit down right now and write a Firefox plug-in, whether or not it is a derivative work of Firefox or not is not affected in any way by whether or not there is some other browser that also accepts Firefox plug-ins. The status of my plug-in depends entirely on the relationship of my code to Firefox's code. If I have copied copyrighted elements (literally or via adaption, translation, etc) from Firefox, mine is a derivative work of Firefix. If I have not copied copyrighted elements from Firefox, my work is not a derived work of Firefox.
A license is free to define whatever terms it wants, and require licensees to obey them. Hence, if I were to use GPL code in my plug-in, I would have to obey GPL, including going along with whatever rules they have on works "derived from" other works, however they define that.
However, if the author of software X is careful to stick within the limits of what copyright allows--he does not copy any copyrightable elements from Y, then the license of Y and whatever terms it defines are irrelevant to Y, at least as far as direct infringement goes. That is, X is safe from Y suing alleging that X violates Y's copyright.
If the people who actually use X and Y together would be violating Y's license, then Y could try to sue X for "contributory infringement". That's when party Z infringes Y's copyright, and party X is held to be responsible for causing or aiding that infringement.
A very important and basic fact about contributory infringement: there can be no contributory infringement without direct infringement. This pretty much rules out contributory infringement when Y is under an open source license, as almost all of these licenses allow Z to do whatever he wants on his own machine, include make derivative works that combine Y with non-free code. No direct infringement by Z means no possibility of contributory infringement by X.
Putting it all together, here's what I would recommend for someone writing some piece of software, X, that is meant to work with some other piece of software, Y, as a plug-in, enhancement, patch, wrapper, etc. Assume Y is under an open source license.
1. Look at the license of Y. Are you willing to use the same license for X? If the answer is yes, use that license. Go code. Don't worry about possibly copying code from Y. (Assuming Y is using an open source license, of course).
2. If you aren't willing to use the same license as Y, are you willing to use a license that is compatible with it? For instance, Y uses GPL and you are willing to use BSD. If yes, use a compatible license. Go code. Do not copy code from Y (directly or via translation, adaptation, etc). Go ahead and look at Y if you need to in order to figure out something.
3. If you are not willing to use a license that is compatible with Y's license, be careful. Do not copy any code from Y, directly or via translation, etc. Avoid looking at Y's code at all. If you need to know something, such as the interface to a function of Y, get it from the documentation. If the documentation isn't adequate, get somebody who is NOT going to work on X's code or design to look at Y and write a spec. Code from that spec.
What you are trying to do in #3 is to limit yourself to doing things that do not require permission from Y's copyright owner. If you can hit that zone, you have no copyright worries, and you can write X and release under whatever license you want.
"However, as a special exception, the source code distributed need not include anything that is normally distributed (in either source or binary form) with the major components (compiler, kernel, and so on) of the operating system on which the executable runs, unless that component itself accompanies the executable."
http://ma.tt/2010/07/syn-thesis-1/#comment-481846 [source]
http://imgur.com/jNs7c.jpg [screen shot]
Chris could definitely make his business work by embracing the GPL rather than fighting it. Cut some sort of deal with Matt, accept code contributions to decrease development time and, generally, use Thesis to launch a million other non-GPL products.
He also seems worried about others stealing/selling his work, this is over-stated. Couldn't he just trademark the Thesis name to mitigate most of this risk? I would also bet that most of his customers would prefer buying the real thing over some unsupported knockoff. GPL or not, people who don't want to pay will find a way to get it for free.