Re: GPL and FSF (was Re: JFS)

[Brett Glass <brett@lariat.org>]

At 03:37 PM 2/8/2000 , Mark A. Hershberger wrote:

>Software put under the GPL is not necessarily owned or controled by
>the FSF.  

At the end of the GPL are detailed instructions saying how to license
code under the GPL. If these instructions are followed, the code
isn't just licensed under the CURRENT version of the GPL but under any
future one the FSF later releases. Yes, thats right: the FSF can change 
the terms at any time. Thus, it effectively gains COMPLETE control of 
how the software can be used.

This practice is deceptive because the suggested language is usually 
adopted as boilerplate by authors who do not understand what it implies.

>Unless the author grants the Copyright to the FSF, 

All authors who contributes to any "GNU" project must grant all
rights to the code to the FSF. So, in fact, this is quite common.
The FSF has ALL rights to GCC, GNOME, etc.

>the
>author still retains control of his software and, as the copyright
>holder, is still allowed to make proprietary derivitives.

The promise of being able to "dual license" is, alas, a trap which
the GPL sets for unwary developers. It seduces developers into 
stamping the GPL onto their code.... But then, once they do so,
they discover that it is virtually impossible to license their work
for money. Here's why.

First, when the GPLed code is released, the author will doubtless
receive suggested changes and improvements -- often in the form of
code. The problem is that, since the initial code was licensed
under the GPL, the code which is contributed back is licensed
EXCLUSIVELY UNDER THE GPL.

Thus, if the original author accepts a single contribution, his 
whole work is irrevocably licensed ONLY under the GPL and his ability
to legally dual-license goes away.

What if the author refuses to accept the changes to avoid this? In
this case, a second mechanism kicks in. The contributor -- or anyone 
else -- can fork the project to create a GPL-only work that competes
directly with the author's and drives him out of business.

Finally, dual licensing does not work because only a very un-savvy
businessman would license code for money when there's a GPLed version
available. This is true for two reasons. First, the existence of the
GPLed version effectively reduces the market value of its functionality
to zero; anyone can get that functionality for free! Thus, if one pays
money to license GPLed code, one is paying for something which has
no market value to end users. This puts the commercial developer "in
the hole" from the start. Second, the GPL provides for a few "loopholes" 
which allow the author's potential licensees to use the code without 
licensing it. (For example, some vendors of print drivers for UNIX invoke 
GNU Ghostscript but then post-process the output through their own software 
after that. They don't change GhostScript itself.) So, in many cases, they 
have no need to license the GPLed code, and the author loses.

Pete Deutsch, the author of GhostScript, fell into this trap and so
resented it that he later penned a different and much more restrictive
license for his product. (Alas, this license has problems of its own:
he chose to make it a more "virulent" version of the GPL which 
contaminates other programs on the same media. It is also probably 
unenforceable due to some vague terms which the author explicitly 
reserves the right to define at his whim.)

While Deutsch at one time he made a reasonable amount of money customizing 
GhostScript for embedded applications, he has not gotten nearly the reward he
would have if he had recognized the traps inherent in the GPL.

>I realize that you may not have been saying that the GPL grants
>ownership to the FSF, but it could have been interpreted that way.

I hope I've clarified things adequately above.

--Brett