kaffe-1.0.b1 probs on Linux

Robert S. Thau rst at ai.mit.edu
Wed Jul 15 11:33:55 PDT 1998

Followups to a few comments on licenses, in an attempt to at least
point out what seem to be a few common areas of confusion: 

Kiyo Inaba writes:

 > Or, even worth.
 > If you use ANY GNU programs with OS's native library which usually
 > is not GPL'ed, it introduces 'linking GPL program with non GPL
 > library'.

Incorrect.  The relevant text from the GPL reads as follows:

  The source code for a work means the preferred form of the work for
  making modifications to it.  For an executable work, complete source
  code means all the source code for all modules it contains, plus any
  associated interface definition files, plus the scripts used to
  control compilation and installation of the executable.  However, as a
  special exception, the source code distributed need not include
  anything that is normally distributed (in either source or binary
  form) with the major components (compiler, kernel, and so on) of the
  operating system on which the executable runs, unless that component
  itself accompanies the executable.

Note the "special exception", which specifically states that you do
not have to distribute the OS's native libraries in any form with
GPLed source.  

Unfortunately, the meaning of this clause is increasingly unclear, as
it becomes increasingly difficult to define what constitutes being
"normally distributed... with the major components... of the operating
system".  Nowadays, anyone can add some proprietary widget to a
modified version of xBSD, or one of the Linux distributions, and
produce an "operating system" which is "normally distributed" with the
widget in question, making it at least apparently legal to distribute
GPLed code which links to the widget (provided that you only
distribute binaries which run on your modified xBSD).  

Of course, if this dodge works, then the "infectious" feature of the
GPL is effectively nullified.  RMS appears to claim that the dodge
doesn't work, but I'm not sure that I've seen a really coherent
argument as to why.

Regardless, one thing that is clear is that linking with the ordinary
C libraries should never be a problem.

One more comment [sorry, I lost the attributions]:

 > > >> You may still include classes.zip *after* Klasses.jar, if you still
 > > >> wish to run `javac', for example, but this is illegal, because
 > > >> classes.zip is not GPL'ed, and you can't link GPL'ed and non-GPL'ed
 > > >> code together.
 > > 
 > > > I think it's only illegal if you plan on distributing it.  Running
 > > > it should be fine.
 > > 
 > > That's not RMS's opinion.

Ummm... could you supply a citation?  I've seen him argue many times
that it is illegal to distribute sources under GPL which are
*intended* to be linked with proprietary libs (or at least, those
which are not "normally distributed... with... the operating system").
This was, in particular, at the center of the GMP/RIPEM flamewar, as
well as the center of a currently ongoing flamewar which I shall not
name, lest it spill over onto this mailing list.

However, I've never seen him claim point-blank that the GPL bars
anyone from linking GPLed code with a proprietary library *for private
use*, so long as the resulting object is *not* redistributed to other
parties.  (IMHO, his somewhat incendiary writing style sometimes leads
people to believe that he is making more expansive claims than a close
reading of his actual words would justify --- but on the other hand,
I've not read everything he's written on the topic).


More information about the kaffe mailing list