The GPL FAQ isn't a legal opinion, and neither is anything I will write, but I think it's bogus.
If the program dynamically links plug-ins, and they make function calls to each other and share data structures, we believe they form a single program, so plug-ins must be treated as extensions to the main program. This means that linking the GPL-covered plug-in with the main program would violate the GPL.
It's pretty trivial to see why. Suppose instead that the DLL was not GPLed, but the main program was. Since they are a single program, that means the DLL must be released under the GPL. Let's say that you can't convince the author to do this. Are you screwed?
No, and let me tell you why. The main program and the DLL are separate works. Merely because we can think of the main program and DLL as having separate authors with separate licenses demonstrates that. You can't make the GPL infect other programs post facto.
References on computers come in all shapes and sizes. From an array index, to a pointer, to a function call, to an object, to a DLL, to a component, to a distributed component, to a database, to a server. It goes on and on. If we take this further, we start questioning whether a web browser like IE can or cannot make a socket connection to a GPLed web server.
Copyright and fair use are meant to deal with quoting, but that's not what's really the intent here. We're dealing with execution, which is more like reading.
Suppose Alice writes a book that refers to a passage in Bob's book. When reading Alice's book, Cathy comes across the reference. Enticed, she puts down Alice's book and reads the long passage in Bob's. Done, she returns to where she left off in Alice's book.
It's ludicrous to think that Alice (or Cathy) violated Bob's copyright. Now, suppose Cathy was severely disabled, so she had a machine that read the book out to her. It also automatically followed references when encountered and read those out. It's still not a violation.
Suppose us lazy people liked this machine, and it became the next craze. Authors begin to write conscious of the machine, and idiot post-modernist wankers even make little games about it. Still violating copyright? Or are we boiling a frog now?
This issue isn't fictional. It's about to become real. It's not going to be long until your browser spiders the links from a page just to provide you with a hint of what's underneath before you waste a click. It's not the fault of the author of the web page that you're reading someone else's work in the middle of hers, it's your fault for reading it that way.
Like I said, computers have lots of references, but it's not the author's problem that the user is following those references--if we take my magic machine as the model.
Before you do, or don't, consider one last argument. Computers are good at following references, but bad at defining boundaries. There's simply no complete way to describe where one piece of software begins or ends. It should be clear that the concept of a file isn't powerful enough. Nor is a process. Nor is even a computer.
Our brains are wired to think of things in the real world, but this ain't there. Even the notions of file, process, and computer are fictions--metaphors--to grab hold of what is fundamentally a slippery infinite tape.
"Look! You're free! Go, and be free!" and everyone hated it for that. --r
[ Parent ]