GPLv3's secretive Additional Terms

[Chris Travers]

On Wed, Apr 21, 2010 at 1:27 PM, John Cowan <cowan at ccil.org> wrote:
> opensource.*.nwo at neverbox.com scripsit:
>
>> GPLv3's possibly most secretive addition is section 7. Additional Terms.
>> The GNU site and the online world in general pretend like it doesn't
>> exist. I'm especially baffed by the statement "when you convey a copy of
>> a covered work, you may at your option remove any additional permissions
>> from that copy."
>
> The key to understanding 7 is to distinguish between additional
> permissions and additional restrictions.  The first are things that you
> want to allow that the GPL would not by itself allow; the second are
> things that you want to forbid that the GPL would not by itself forbid.

Unfortunately, there are a lot of things which aren't entirely black
and white here.  IANAL, etc.

For example, suppose I incorporate a BSD-licensed file into a GPL
v3-licensed work.  Pretty much everyone agrees I can do this, or even
if they don't they at least concede that this was a goal of the GPL
v3.

However, you run into a problem here trying to read the BSD license as
basically the GPL v3 plus additional permissions.  In essence, I don't
think you can simply chance the copyright on the otherwise-unchanged
BSD-licensed file by invoking section 7 of the GPL v3 (and Richard
Fontana from the SFLC has, in the past, opined as much on this list),
but that relicensing restriction is not in line with the additional
restrictions section either.  A very strict reading of section 7 of
the GPL v3 would therefore render it incompatible with the BSD
license.

But this ends up creating a great deal of difficulty.  If the
restriction that, absent significant changes, a file cannot be
relicensed is an implied exception, does that mean that I can include
proprietary software libraries in closely tied ways, creating an
implied additional restriction as well?  To be honest, even after a
few years, I still don't have a principled way of reading this
session.  I think the closest one could come to would be a
"substantive interest" test where additional restrictions might be
allowed if they didn't prejudice a substantive interest on the part of
the licensor, such as source code availability, or the ability to
refactor code.  This would allow open source licenses generally to be
deemed compatible (perhaps to a much greater extent than the GPL v2--
would the MPL be compatible?) but might stop short of limiting it to
files used.

I don't know though.  This (and the AGPL compatibility exception
written in) are major reasons why I avoid releasing work under this
license.

I wouldn't call it secretive though.  Just that compatibility in
licensing is a difficult thing and that there are some important
shortcomings there.

Best Wishes,
Chris Travers