GPLv2 enforcement

Posted Mar 22, 2007 9:09 UTC (Thu) by malor (guest, #2973)
In reply to: GPLv2 enforcement by bug1
Parent article: The Torvalds Transcript (InformationWeek)

You're clearly and unambiguously wrong about this.

Say, oh, Broadcom releases Windows drivers for their wireless chips. They work nicely in Windows.

Then the LinuxAnt project comes along and writes a wrapper that allows you to load the the Broadcom driver under Linux. That driver knows nothing about Linux; it thinks it's being loaded under Windows.

LinuxAnt is required to release their source code, because their module is obviously a derived work of the kernel. They are not, however, required to do so with the Broadcom source code, and neither is Broadcom. It wasn't developed for Linux and knows nothing about it, so it cannot be a derived work. It is really that simple.

By your argument, if we write a special wrapper to let Windows itself run under Linux, suddenly Windows becomes a derived work and Microsoft loses the ability to sell it without source code.

You're really trying to go after NVidia's binary blob approach here. I think they are immune; it's my understanding that their binary is just the Windows driver, and they wrap it in a glue layer. That doesn't make the binary blob a derived work.

to post comments

GPLv2 enforcement

Posted Mar 22, 2007 11:00 UTC (Thu) by Wol (subscriber, #4433) [Link]

Plus it's only a copyright infringement by the DISTRIBUTOR.

If the distribution downloads the blob from NVidia on installation, who's breaking copyright? Not the distro, they never mixed GPL and proprietary. Not NVidia, they own it. And not the user, they had NVidia's permission.

It is PERFECTLY LEGAL to mix GPL and proprietary code. You just can't distribute the result. If I wanted to ship a program that was a mix of proprietary and GPL, I would simply ship a build environment. All the GPL as source, the proprietary as object modules. Get the user to run a make script to build the program. ALL LEGAL AND ABOVE BOARD.

It just wouldn't look good to a user, and I wouldn't do it because I don't think it's in the spirit of the GPL. But it's perfectly legal ...

Cheers,
Wol

GPLv2 enforcement

Posted Mar 22, 2007 11:00 UTC (Thu) by thebluesgnr (guest, #37963) [Link]

If LinuxAnt distributes only the GPL loader and nothing more it is obviously not violating the license.

The problem is distributing the loader and proprietary bits together, as nVidia does.

Also, take for example a distribution that ships such driver with the kernel. They distribute a version of Linux that "can accelerate nVidia cards", which requires modifications to Linux which they don't make fully available. IMO they're clearly violating the GPL.

GPLv2 enforcement

Posted Mar 22, 2007 11:07 UTC (Thu) by bug1 (guest, #7097) [Link]

According to the US gov, a derivative work is, "a work that is based on (or derived from) one or more already existing works" (1)

Could the LinuxAnt wrapper (compatibility layer) be written without the windows broadcom driver, without the broadcom driver does the compatibility layer have a purpose ?

I think most real people can see there has to be some sort of dependency there... But maybe the compatibility layers dependence is through a standard interfaces, which are allowed exceptions in most countries i believe. (but im not sure how that effects the GPLv2).

You misunderstand me in your windows example, what im claiming is that the compatibility layer is the derivative work of both layers, the GPL requires athe source of dependencies be distributable. There are no requirements placed on either layer unless they want to be legally compatible with the compatibility layer.

If Microsoft doesnt distribute this hypothetical microsoft/linux compatibility layer they dont have to do anything.

But its all depends on the interfaces, if it uses standard interfaces to windows then the compatibility layer is probably ok, but as you suggest in your first post, only a judge can say for sure.

Another example would be shell script, it serves no purpose without a shell, so it is dependent on _a_ shell, but the shell is a standard interface (POSIX SUS), there are any number of shells that can interpret it, so the script isnt a dependency on any specific shell.

Can the linuxAnt compatibility layer claim the same sort of independence from the windows broadcom driver, does it access the driver directly, or _only_ through standard windows calls ?