Is it appropriate to include a GPLv3 library into a GPLv2 work? (A work that is expressly *not* GPLv2+?)
If not, why? (The "for any purpose" clause seems to cover this from both directions -- and anyway they are both GPL licenses so this is definitely within the spirit and intent of the GPL...?)
The specific case is LibreCAD (forked from GPLv2-only QCad CE) and OpenDWG (a GNU project which is GPLv3-only). We have communicated with the OpenDWG maintainer and he's said a few times that he would add a "GPLv2 or later" clause so that we could include the code without any confusion, but this has not yet happened for whatever reason. We have not yet fully refactored the QCad code so it is still truly a fork and not an independent work, so we cannot upgrade the license. The original author from Ribbonsoft has written us to express his intent to block us from upgrading the license for as long as possible (he is also now trying to claim that the manual included with QCad CE is not open source -- but this is a different issue).
My understanding of the precise limits of the GPL and LGPL is a little fuzzy when it comes to the way routines are included in different GPL versions. It may be that the rule itself is a bit fuzzy. We know that the OpenDWG folks wouldn't have any objection, but we're also not looking to have any outside party suddenly become adversarial and start pointing fingers ("that's ridiculous!" won't stop someone from trying).
If someone has a resource that covers the ins and outs of mixing GPLv2, 2+ and 3 clauses I would be most appreciative. We're a bit confused about: 1- Including new source files (our own) as GPLv3 vs GPLv2+ within a project that is, as a whole, considered GPLv2 2- Linking GPLv3 from old GPLv2 code (such as the OpenDWG situation) 3- The point when a project is considered to be completely reworked and no longer the property of the original author (such as what happened with parts of the SCO case a while back). What is the threshold?
zxq9 wrote:
If someone has a resource that covers the ins and outs of mixing GPLv2, 2+ and 3 clauses I would be most appreciative.
That's what this matrix is for: http://www.gnu.org/licenses/gpl-faq.html#AllCompatibility
This chart may also help: http://www.gnu.org/licenses/quick-guide-gplv3.html#new-compatible-licenses
Björn Persson
On 09/25/2011 04:13 AM, Björn Persson wrote:
zxq9 wrote:
If someone has a resource that covers the ins and outs of mixing GPLv2, 2+ and 3 clauses I would be most appreciative.
That's what this matrix is for: http://www.gnu.org/licenses/gpl-faq.html#AllCompatibility
This chart may also help: http://www.gnu.org/licenses/quick-guide-gplv3.html#new-compatible-licenses
Thank you. Very helpful. Also indicates that OpenDWG won't wind up anywhere meaningful for quite a while.
On Sun, 2011-09-25 at 03:32 +0900, zxq9 wrote:
Is it appropriate to include a GPLv3 library into a GPLv2 work? (A work that is expressly *not* GPLv2+?)
If not, why? (The "for any purpose" clause seems to cover this from both directions -- and anyway they are both GPL licenses so this is definitely within the spirit and intent of the GPL...?)
<tangent> The relevant intent here is that of the projects that chose to grant GPLvN only rather than GPLvN+. FSF would certainly prefer if everyone granted GPLvN+ (Savannah even requires it), but understandably, some projects do not trust FSF as a steward of future license upgrades. For these cases, the GPL (unlike some other free software licenses) is designed to make it possible to grant GPLvN only. Had that option not been available, we would likely have seen further fragmentation of licenses.
So, actually, it directly contravenes the intent of the projects to combine GPLv2-only and GPLv3-only work. (Let's leave aside the question of the tenability of FSF's claim that dynamic linking creates a derivative work.) </tangent>
-
Björn Persson -
Matt McCutchen -
zxq9