Circumventing the GPL via open standards

On Fri, 7 Jun 2002, Patrik Stridvall wrote:

> Sure, but you can write a novel that is pretty close to another
> novel without being a true derivation. There is court case
> giving precendence in that area.
> 
> Again note that copyright only protects unique expression and there
> is usually quite a lot more unique expression in a novel than 
> your run of the mill GPL:ed software which in large parts lies
> much closer to the facts, idea, methods etc that are unprotected
> by copyright
> 
> So I think it would in most cases be quite easy to avoid being
> a derivation. Especially if the GPL:ed work is used unmodified 
> in a larger proprietary work. The GPL:ed usually doesn't allow
> this so it would effectively circumvent the GPL.

I think this is something Stallman fears.  The GPL can be circumvented by 
open standards.  As long as it's not legally a "derivative work" under 
copyright law, the GPL has no force whatsoever.

For example, consider an open standard, such as Project UDI, which was 
developed independently of any GPL code.  Clearly, the GPL holds no claim 
to this standard.  UDI stands for "Uniform Driver Interface", a low-level 
interface, source-code compatible and binary-compatible where possible, for 
many operating systems to be able to use the same driver code unmodified.
The standard places restrictions on the OS and the driver, but it's just an 
API definition, and it was developed using proprietary operating systems.

Now, suppose Microsoft (or any other vendor) implements the UDI operating 
system ("environment") specification.  This implementation could be said to 
be a derived work of the UDI standard (and even that is open to question), 
but certainly not a derived work of anything covered by the GPL.  Microsoft 
could add UDI support to Windows if they wanted to.  (But they won't since 
it weakens their monopoly.)

Now, suppose someone (maybe even Microsoft?) pulls GPL'd drivers out of the 
Linux kernel and ports them to conform to the UDI interface.  Anyone could 
do this, and the GPL requires that they release the source code to go with 
that binary.  Fine so far.

Now, what happens if Microsoft Windows has a UDI environment, and the user 
dynamically loads a UDI driver, even if the driver is under the GPL?  Does 
the entire OS suddenly become a derivative work?  Obviously, this is a 
ridiculous conclusion, especially if Microsoft implemented UDI without any 
intent to use GPL'd drivers.  There's no way the courts would allow a mere 
user to obligate a vendor in this way.  Therefore, it surely would not be 
considered a derivative work.  And if it's not, the GPL doesn't apply.

Does it become a derivative work if Microsoft does the porting?  That's 
more of a gray area perhaps, but it still seems unlikely.  What if they 
include the dynamically-loadable GPL'd UDI drivers (and source) on the 
Windows distribution media?  Is that "mere aggregation" or does it become a 
derived work at that point?

At what point is it a derived work?  It's not clear.  But it's clear that 
situations could arise where a proprietary product is linked with GPL'd 
code via a neutral API without the knowledge of the proprietary author, and 
there's no way the GPL could be held to bind the proprietary author then...

Deven