On Tue, May 11, 2004 at 05:23:48PM -0400, Raul Miller wrote: > > > > As a rough idea, imagine if gcc were made to support > > > > special keywords or control files to make it easier to build > > > > programs which use palladium's proprietary encryption and > > > > digital rights management facilities object model. Or, > > > > more generally, imagine any change which makes gcc into > > > > something that works in a proprietary fashion.
On Tue, May 11, 2004 at 05:37:51PM -0400, Glenn Maynard wrote: > > > This is allowed by the GPL and required to be allowed by the > > > DFSG, of course, as long as the resulting gcc binary can be > > > distributed under the terms of the GPL. The GPL doesn't care > > > what kinds of changes you make (with very limited exceptions, > > > such as the license blurb). On Tue, May 11, 2004 at 07:32:23PM -0400, Raul Miller wrote: > > Only if the resulting work (including the implementation of the > > support for those keywords) is distributable under the terms > > of the GPL. On Tue, May 11, 2004 at 08:42:08PM -0400, Glenn Maynard wrote: > "The license must allow modifications and derived works, and must allow > them to be distributed under the same terms as the license of the > original software." Yes. > DFSG#3 requires that derived works be distributable under the terms of > the original software. Yes. > The GPL requires that derived works *only* be distributed under its terms > (with no further restrictions, hence its notorious virality). Yes. > If the proprietary code is not under the terms of the GPL, then the GPL > prohibits distribution, because it is *not* under the same terms as the > license of the original software. The part it prohibits distribution of is the part that is under the original terms -- the GPL has no say over the parts under licensed under other terms. This includes the work as a whole, however, since the work as a whole must satisfy all licenses. > This is a case in which DFSG#3 very explicitly does not require derived > works to be distributable. Agreed, though the convolutions you went through don't really have any bearing on this point. > It only requires that derived works be distributable under the terms > of the original software. Yes. -- Raul