[kolla]is it OK to modify python code in OpenStack project?
Recently, Kolla project has requirement to modify[1] the python's ConfigParser.py code[0].
Python is using PSF license[3], which is GPL compatible. But OpenStack is using Apache License.
Here is the diff view[2].
I want to make sure: is it OK to re-license ConfigParser.py? If not, what the solution?
[0] https://github.com/python/cpython/blob/2.7/Lib/ConfigParser.py [1] https://review.openstack.org/412101 [2] https://gist.github.com/jeffrey4l/2258b276cbd038e73797cfa0952da371/revisions... [3] https://docs.python.org/3/license.html
Jeffrey Zhang wrote:
Recently, Kolla project has requirement to modify[1] the python's ConfigParser.py code[0].
Python is using PSF license[3], which is GPL compatible. But OpenStack is using Apache License.
Here is the diff view[2].
I want to make sure: is it OK to re-license ConfigParser.py? If not, what the solution?
[0] https://github.com/python/cpython/blob/2.7/Lib/ConfigParser.py [1] https://review.openstack.org/412101 [2] https://gist.github.com/jeffrey4l/2258b276cbd038e73797cfa0952da371/revisions... [3] https://docs.python.org/3/license.html
I'm not a lawyer, but it sounds slightly tricky.
The PSF license is Apache-compatible, which means that you can combine PSF-licensed and Apache-licensed code (if you retain the original licenses). But OpenStack code must be licensed under a license supported by the Contributor License Agreement (CLA) which allows redistribution by the OpenStack Foundation under ASLv2 (currently only ASLv2, the MIT and both forms of the BSD license meet this requirement).
So I don't think we can really bundle PSF-licensed code within OpenStack code, and I think too much was copied from the original class so that anyone can pretend it is original work.
Three solutions that should work around the problem without requiring lawyers to further investigate the issue:
1/ Do a clean room implementation (by someone not familiar with the original code)
2/ Rewrite your code so that it takes a pure ConfigParser class output and transforms it into what you need, rather than copying code over
3/ Publish your modified ConfigParser into its own 3rd-party PyPI library (licensed under the PSF license)
Of those, (2) is the one that introduces the less technical debt (no code copy).
Thierry is correct in his analysis.
If you really want to copy it - which I don't suggest - then copy it over into your repo, rename both the module and the class, and make your code use your derived work.
Put a header at the top that says something like "Based on Python's ConfigParser.py. Original ConfigParser.py distributed under the Python Software Foundation license. Additions and changes copyright [YOU] and distributed under the Apache Software License, v2."
Do not modify the stdlib in-place (as it sounds like you are doing) - but perhaps I just misunderstand.
________________________________ From: Thierry Carrez thierry@openstack.org Sent: Monday, January 2, 2017 4:45 AM To: legal-discuss@lists.openstack.org Subject: Re: [legal-discuss] [kolla]is it OK to modify python code in OpenStack project?
Jeffrey Zhang wrote:
Recently, Kolla project has requirement to modify[1] the python's ConfigParser.py code[0].
Python is using PSF license[3], which is GPL compatible. But OpenStack is using Apache License.
Here is the diff view[2].
I want to make sure: is it OK to re-license ConfigParser.py? If not, what the solution?
[0] https://github.com/python/cpython/blob/2.7/Lib/ConfigParser.py [1] https://review.openstack.org/412101 [2] https://gist.github.com/jeffrey4l/2258b276cbd038e73797cfa0952da371/revisions... [3] https://docs.python.org/3/license.html
I'm not a lawyer, but it sounds slightly tricky.
The PSF license is Apache-compatible, which means that you can combine PSF-licensed and Apache-licensed code (if you retain the original licenses). But OpenStack code must be licensed under a license supported by the Contributor License Agreement (CLA) which allows redistribution by the OpenStack Foundation under ASLv2 (currently only ASLv2, the MIT and both forms of the BSD license meet this requirement).
So I don't think we can really bundle PSF-licensed code within OpenStack code, and I think too much was copied from the original class so that anyone can pretend it is original work.
Three solutions that should work around the problem without requiring lawyers to further investigate the issue:
1/ Do a clean room implementation (by someone not familiar with the original code)
2/ Rewrite your code so that it takes a pure ConfigParser class output and transforms it into what you need, rather than copying code over
3/ Publish your modified ConfigParser into its own 3rd-party PyPI library (licensed under the PSF license)
Of those, (2) is the one that introduces the less technical debt (no code copy).
-- Thierry Carrez (ttx)
_______________________________________________ legal-discuss mailing list legal-discuss@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/legal-discuss
participants (3)
-
Jeffrey Zhang
-
Thierry Carrez
-
Van Lindberg