invalidate_mmap_range() again [LWN.net]

invalidate_mmap_range() again

Posted Feb 26, 2004 12:24 UTC (Thu) by ekj (guest, #1524)
In reply to: invalidate_mmap_range() again by simlo
Parent article: invalidate_mmap_range() again

There's a fairly fundamental difference;

IBM is going to Linus and saying the equivalent of: "Please make these changes to your code, so that my proprietary kernel-module will work under Linux."

As far as I know, noone has ever approached SCO and asked them to change their code in order to make something or other involving Linux or free software in general possible.

A program that can be compiled and run under any of half a dozen oses is obviously not a derivative of any of them.

A module that is hacked up in a way (with a bsd compatibility layer and whatnot) that it can ONLY work with one spesific OS (Linux) and ONLY if that OS chooses to incorporate changes making the module run is a lot more intimately tied to the OS.

(Log in to post comments)

invalidate_mmap_range() again

Posted Feb 26, 2004 12:59 UTC (Thu) by simlo (subscriber, #10866) [Link]

There are morally differences. But even though the interface is changed and the system can only run on Linux the filesystem itself is not anymore a derivative work of Linux than RCU is a derivative of Unix.

Ofcourse Linux + the filesystem is a kind derivative work as if Linux + Nvidia driver. If distributing <Linux>+<Nvidia driver> is legal then is
<your own version of Linux>+<what ever file system module you want to make>.

You can't have legal decitions based on "how integrated it is", "if it was
written for other os'es." It must be either or. And in the end a judge must decide on where the line is.

The decision I would want was that you can do anything to the kernel you want - including changes to the interfaces. You then have to destribute that under GPL. But on that product you can load whatever module you want.
If that can't be seen as within the GPL, no distributing binary modules along with the kernel is within the GPL. I can't see that a judge can decide "in between".

I wish we could have a case going to court such we can get this decided. The worst thing is that we don't know what is what. If it is decided that binary modules isn't ok, we can always make userspace drivers. I am pretty sure that in that case RedHat will put such an interface into their Linux even if Linus wont because people _need_ non-GPL drivers (to have them in userspace would also greatly impact stability.)

IBM's not going to Linus

Posted Feb 27, 2004 16:57 UTC (Fri) by giraffedata (subscriber, #1954) [Link]

IBM is going to Linus and saying the equivalent of: "Please make these changes to your code ...

I need to correct a significant misconception here: IBM is not going to Linus. IBM is going to (the addressee's of Paul McKenney's email) the nebulous Linux kernel community, the nebulous Linux 2.6 kernel community, and Andrew Morton. And it (the Linux kernel) is not Linus' code.

I don't believe it was mentioned in the articles whose decision it was to include the original EXPORT_SYMBOL_GPL into the Linus/Andrew tree or whose decision it was to change it to EXPORT_SYMBOL, or if one or the other other was a complete accident.