On Fri, Jun 14, 2002 at 11:00:46AM +0300, Richard Braakman wrote: > But look at the _next_ paragraph: > > > But when you distribute the same sections as part of a whole > > which is a work based on the Program, the distribution of the > > whole must be on the terms of this License, whose permissions > > for other licensees extend to the entire whole, and thus to each > > and every part regardless of who wrote it. > > The GPL grants "permissions for other licensees", which include deleting > arbitrary parts and distributing the result. If the GDCL does not grant > this permission, then it is not compatible. (What's the "GDCL"?) But look at the qualifying clause: But when you distribute the same sections as part of a whole which is a work based on the Program, ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ That means the chunks that are GPLed as part of the Program. If somebody extracts out the "identifiable sections of that work are not derived from the Program," and bases a work on that, the GPL doesn't apply. Whatever license was originally on those identifiable sections does. Consider the GPL FAQ: If a program combines public-domain code with GPL-covered code, can I take the public-domain part and use it as public domain code? You can do that, if you can figure out which part is the public domain part and separate it from the rest. If code was put in the public domain by its developer, it is in the public domain no matter where it has been. The same logically follows for any code which is under copyright but licensed under GPL-compatible terms. If the DFCL grants permission for its normally GPL-incompatible requirements to be ignored when a working using the DFCL is combined with a GPLed work, then it's not GPL-incompatible. It's a virtual dual license; there's a switch that gets thrown when you integrate a DFCLed work into a GPLed one. I assert that to state that "given two *identical* works, if one bears a GPL license and the other an MIT license, that the MIT licensed one *must* be illegitimate because the same work was under the GPL because at some point" is patent nonsense. If it is nonsense for unmodified works, it is also nonsense for modified works that remain under the original copyright as well. The GNU GPL is not fairy dust. You cannot sprinkle it over a work that is under some other license, and change that license. You cannot sprinkle it over a work that is in the public domain, and take that work out of the public domain. > I think that your arguments about what can be done with combined works > are coming from the wrong side of this. The question is whether the > GPLification clause of the DFCL is sufficient, I assert that it is, and I offer the GNU GPL text and the FSF's GPL FAQ as support. > Your Windows 2000 example illustrates this wrong-sidedness best: > > > Here's a question for you. If I get my mitts on the source code to > > Windows 2000, and change one line in a source file, can I then > > distribute that source file under any license I want? The issue isn't > > the GPL, it's whether copyright is preserved in a work despite its > > modification or aggregation with a GPLed work. The answer, as I > > understand it, is "of course the copyright is preserved." > > The answer is "no", because you don't have a license for Windows 2000 > which grants redistribution under any terms. Permission to redistribute is not permission to relicense. You are reinforcing my point, not undermining it. > If you did have a license for it that is compatible with the GPL, then > the answer would be "yes". The GPL certainly does not give me permission to relicense a GPLed work I have modified under any license I want. > (I think this is in fact the test of GPL compatibility.) You are missing the point. This isn't really about the GPL at all, but whether the identity of a copyrighted work is a meaningful concept. (Given that copyright is in fact a legal fiction, it's not surprising that there is much disagreement and confusion about.) > > It doesn't *matter* under copyright law (as I understand it) whether I > > passed a DFCLed document through a GPLed software strainer. If what > > comes out is a derivative only of the DFCLed document, only the DFCL > > applies. If what comes out is a derivative of both the DFCLed original > > document and the GPLed software, then each licenses applies to various > > parts of the work, depending on the circumstances. > > But the GPL claims to grant its permissions to the whole, including every > part regardless of who wrote it. That's correct, and the DFCL does not contest the terms of the GPL, when it a DFCLed work is living within the context of the GPL. > If that statement in the GPL is not true, then you have not > (effectively) applied the GPL. These requirements apply to the modified work as a whole. In other words, they do not necessarily always apply to every single bit of a modified work. If identifiable sections of that work are not derived from the Program, and can be reasonably considered independent and separate works in themselves, then this License, and its terms, do not apply to those sections when you distribute them as separate works. Let us try to work through this in practice with an easy case. I record a WAVE file of myself playing the electric guitar and release it under the DFCL. Some extremely demented person wants to inline this WAVE file into GNU Emacs such that it plays when Emacs starts up. This demented person doesn't understand startup hooks or elisp, so he hacks on Emacs's C core and inlines the whole damn WAVE file as hex constants inside a huge data structure.[1] I assert that my guitar noodling is: 1) identifiable 2) not derived from the Program 3) can reasonably be considered an independent and separate work from GNU Emacs Therefore: The GNU GPL, and its terms, do not apply to my lousy guitar performance when it is distributed as a separate work from GNU Emacs -- even if someone distrubutes it as a C file with all the GNU Emacs code ripped out, and leaving just the huge data structure containing hex constants. But when you distribute the same sections as part of a whole which is a work based on the Program, the distribution of the whole must be on the terms of this License, whose permissions for other licensees extend to the entire whole, and thus to each and every part regardless of who wrote it. This demented person packages and distributes his forked version of GNU Emacs as Guitarmacs. I assert that: 1) WHEN Guitarmacs is distributed, it contains my independently copyrighted and licensed guitar performance. 2) Guitarmacs contains the same sections that were separately distributed under the DFCL. 3) Guitarmacs is based on GNU Emacs. 4) The distribution of Guitarmacs is required to be under the terms of the GNU GPL. 5) This includes the parts that are my guitar performance. Therefore: The GNU GPL, and its terms, apply to my lousy guitar performace when it is distributed as part of a work that is a deriviative both of my performance and GNU Emacs, that is when it is distributed as Guitarmacs. When who imported, merged, split out, transformed, or did whatever to whichever hunk of data is irrelevant. Summing up: 1) The GNU GPL, and its terms, do not apply to a DFCLed work when it is distributed as a separate work in itself from a GPLed work. 2) The GNU GPL and its terms do apply to a DFCLed work when it is distributed as part of a work that is derived from a GPLed work[2]. The only ambiguities here are the meanings of "work" and "derivation". These ambiguities are not unqiue to the DFCL or even the GPL, and the DFCL licensing enterprise doesn't really have any bearing on them. [1] I could have some of these details wrong; I use vim. [2] You must necessarily derive someone's work to get from a GPLed original to one that includes someone else's work, even the GPLed work is your own, and you are authoring it for the sole purpose of incorporating someone else's work. There is still a part of the total that is yours; that is your work, and the part to which the GNU GPL is directly applied. If there is no GPLed work to derive from, then the GPL doesn't enter the picture and this whole discussion is irrelevant. :) -- G. Branden Robinson | You could wire up a dead rat to a Debian GNU/Linux | DIMM socket and the PC BIOS memory branden@debian.org | test would pass it just fine. http://people.debian.org/~branden/ | -- Ethan Benson
Attachment:
pgpUTJY7Sdn0Y.pgp
Description: PGP signature