🚀 go-pugleaf

OK, here is something for all you self-appointed software lawyers to
fight over... :-)

assumptions:
	1) If you make a binary out of of source code, where one part of the
           source code falls under the GNU-license, and distribute that
	   binary, all sources must fall under GNU-license.
	2) If you use a GNU-licensed header file, you are using GNU-licensed
           source code.
conclusion:
	3) If you distribute a binary of emacs or gcc, you are breaking the
	   GNU-license.

explanation:

1) is my interpretation of the COPYING file. This has also been said
many times on gnu.misc.discuss - so many times, in fact, that I almost
tend to believe it :-). This is the famous so called "virus clause".

2) This is something that also has been said many times on
gnu.misc.discuss as a counter to the "If I just distribute
object-files and let the customer link" GNU-license loophole claim. 

This also seems rather reasonable: Include files are a rather major
part of the program code, especially if you use macros a lot.

Here comes the big *BUT*:

BUT, if this works one way - doesn't it have to work the other way
too?  If the include file is such a major part of the software that
inclusion of it makes it it fall under the GNU-license, how about all
the other includes? For example:

	#include <stdio.h>
	#include <string.h>

Reasonably, there should not be any difference between files you
include with '<>' and '""'. There should also be no difference between

A)	#include "/usr/include/stdio.h"

and

B)	#include "/usr/local/gnu-code/tree.h"

The stdio.h code in A) is just as much a part of the total program as the
gnu code in B), isn't it??
	
That means: *The stdio.h code must also fall under GPL if it is used in a
	     GPL source code!!!*

BUT is there any system where stdio.h is yours to place under GPL??? I
don't think SUN would be very happy if you "gave away" their stdio.h
by distributing a binary of gcc for the Sparc...

---
Exactly the same argument may be made with:
	Making a library out of source code under gnu-license (but not
	under the library license) instead of "tree.h"
and:
	The propraritary libc.a on the system where the binary was made.
	(Assume it is a system without dynamic libraries.)
But I'll save you the doubling of the article length... :-)

[And if you say "Oh yes, but libc.a is *different*": What is stopping
 me from inserting gnu code into my libc.a just when I compile my
 propraritary binary? Copyright law says nothing about what is the
 "proper" contents of libc.a...]  
---

Seems we have to wait for the *total, self-contained GNU system* until
we can distribute GNU binaries. Anyone who wants to sue Next for the
source code to their libc.a? After all, its header files probably are
part of the source code of their gcc... That would be a nice court
precedent to resolve that gray area of "how far can the GPL go" :-)