samba license change

[Matthew Saltzman]

On Wed, 2007-10-10 at 23:41 -0400, Simo Sorce wrote:
> On Wed, 2007-10-10 at 18:57 -0500, Chris Adams wrote:
> > Once upon a time, Simo Sorce <ssorce at redhat.com> said:
> > > On Wed, 2007-10-10 at 13:21 -0800, Jeff Spaleta wrote:
> > > > But Nicolas has stated my murkier concern. If we just drop re-licensed
> > > > libsmbclient into place with no enforced technical break like a soname
> > > > change or a library renaming, are we acting negligently with regard to
> > > > protecting our own users who consume pieces of rawhide to suppliment
> > > > F7 or soon to be F8? If the re-licensed code can just drop into place,
> > > > are we encouraging users to violate the license at runtime by making
> > > > it too easy to use the re-licensing binary in situations where its
> > > > inappropriate?
> > > 
> > > Short answer: no
> > 
> > So what if KDE used a private copy of libsmclient.so.0 (from 3.0 so
> > GPLv2) during build but didn't ship it?  As long as the newer version is
> > ABI compatible (even if the license isn't), is there a violation?  I'm
> > sure some would say it is a violation of the spirit, but is there a
> > technical violation?
> > 
> > If so, where is the violation?  Where does it physically occur?
> > 
> > I haven't read GPLv3 closely, but "linking" doesn't appear anywhere in
> > GPLv2 except in a single note in the appendix (not in the license terms
> > itself).  Everything else in GPLv2 talks about derivative works.
> > Arguably, KDE's use of libsmbclient is a derivative work of a GPLv2
> > product.  As long as the interface to that product doesn't change (so
> > KDE's use of the product doesn't change), KDE could claim to be a
> > derivative work of the GPLv2 product.  Samba is NOT changing the ABI, so
> > any derivative works (that are derivative by action of using the
> > libsmbclient.so.0 interface) can still claim to be derivatives of only
> > the GPLv2 library.
> > 
> > I'm really curious about this (not just trying to still the flames): if
> > a GPLv2-only program is linked against a GPLv2 (or 2+) library and the
> > library switches to GPLv3 (or 3+), who is violating the license?  Use
> > (e.g. an end-user actually loading the KDE binary that dynamically links
> > against libsmbclient.so.0) is not covered by the GPLv2, so the end-user
> > is not violating it (because they aren't distributing).  A distributor
> > could be building against a GPLv2 version of the library but only
> > distributing the GPLv3 version; is that a violation (why)?
> 
> Far from being a flame.
> Your question is perfectly legitimate, but I don't think there is
> anybody on this list with enough legal expertise to answer.
> We need legal advice to be able to answer this question.

Probably a good idea.  But one point to consider is: In GPLv2, there is
a clause that says that you do not need to GPL a work that links to a
GPLv2 library if there is another library that performs the same
function that you could (in principle) link to instead.  If there is a
similar clause in GPLv3, that would seem to make the situation Chris
describes not a violation.  It seem that as long as the ABIs remain
identical, the existence of a GPLv2 lib that one could link to would
mean that code linking to the GPLv3 lib would not have to be GPLv3.

IANAL, of course, but it sounds at least like a plausible interpretation
to me.

> 
> Simo.
> 
> 
-- 
                Matthew Saltzman

Clemson University Math Sciences
mjs AT clemson DOT edu
http://www.math.clemson.edu/~mjs