libgcj / Classpath relicensing and cooperation

Sat Apr 1 00:00:00 GMT 2000

On Wed, Mar 08, 2000 at 03:39:49PM -0800, Per Bothner wrote:
> Paul Fisher <rao@gnu.org> writes:
>> > >From the Kaffe licensing FAQ:
> > 
> > Can I run proprietary Java applications and native JNI libraries
> > under Kaffe?
> > 
> > Yes, you can. Kaffe's choice of GPL does not affect your ability
> > to run any Java or JNI-based code that you could run on any other
> > Java virtual machine.
> > 
> > The same would apply to running Java applications under the GPL'd
> > Classpath AWT.
>> Unfortunately, an FAQ is not a legal document. The words of the GPL
> do not support this interpretation. More to the point, it seems to
> contradict how the GPL has traditionally been interpreted, including
> historical pronuncements from the FSF. What seems to be intended is a
> "GPL with modifications" - i.e. more-or-less what Cygnus has been
> proposing all along.

I don't often speak up on this list, and I have no desire to start a
license flamewar. However, for the benefit of those less familar with 
this territory, a couple of points. The GPL could only conceivably
apply in this case of the mere execution of a program under kaffe
constitutes the creation of a derived work (of kaffe). 
This is the grey area in which dynamic linking resides. An
interpretation which convinces me (and I believe that RMS agrees with
me in broad brush strokes, if not in some of the details) is that a
program always intended to dynamically link against libXYZ (where XYZ
is GPL) is essentially a derived work of libXYZ (by considering the
parallel with static linking, and the fact that the difference is a
mere technicality irrelevant to law). However, I'd say that a program 
compiled against a generic interface and quite capable of dynamically
linking against libXYZ or libABC is a derived work of
neither. (interfaces are not copyrightable) (this is the argument by
which the mere creation of a working harmony project would make KDE ok,
even though KDE continues to run against Qt)
Similarly, I'd argue that an (evil) java program which in some
underhand way used some proprietry kaffe VM features might be argued
to be a derivative work of kaffe, but a java program which can be run
under any VM is not a derivative work of kaffe, irrespective of the
fact it *can* be run under kaffe.
Replace kaffe with libgcj, the paragraph still makes sense ;-)
You don't have to agree. If you represent a company, consult a lawyer 
and don't just follow my advice. Think before flaming (I've been down
this path too often...)
Jules
-- 
Jules Bean | Any sufficiently advanced 
jules@{debian.org,jellybean.co.uk} | technology is indistinguishable
jmlb2@hermes.cam.ac.uk | from a perl script