License

[fakerybakery]

Hi,
Amazing package! Would you consider switching to a more permissive license such as MIT, ISC, or LGPL so all upstream packages don't also have to be GPL-licensed?
Thank you!

[maxbachmann]

Most of the library is MIT licensed, since it directly uses the implementations from RapidFuzz. You can directly use https://github.com/maxbachmann/RapidFuzz for these parts as well. Possibly this could be documented to make clear that parts of the library are licensed under the MIT license. The only parts which are under a GPL license are:

• Levenshtein.quickmedian
• Levenshtein.median
• Levenshtein.median_improve
• Levenshtein.setratio
• Levenshtein.seqratio

These implementations got written by the original author of the library and so I can't simply change their license. To change their license I would need to rewrite these algorithms from scratch. To my knowledge these algorithms are not used a lot, since as the name suggests most people use the library for levenshtein calculations. I might still rewrite them at some point, but I have a lot of other things to do in my open source projects and so this might take quite a while.

AFAIK the biggest part in a rewrite would be finding something like a paper describing these algorithms that could be used as basis for a new clean room implementation.

[Joerki]

Hi @maxbachmann,

if I see it correctly the license declaration of GPL-2.0 in LICENSE file (GPL-2.0-only) and COPYING of the final package based on the cpp source is not correct.

According to Readme.md and cpp file also the LICENSE file and licenses/COPYING in the distributed package has to reflect that GPL-2.0-or-later is applicable.
The latest entry in pyproject.toml is correct on the other hand, so the metadata (on pypi and pip show) look fine.

See:
https://spdx.org/licenses/GPL-2.0-or-later.html

Another point is that I did not see a copyright notice for David Necas in the distributed package.
But it is an obligation to clearly specify the original copyright holder in the package you distribute in a transparent way e.g. the license file.
Mentioning him in authors list of pyproject.toml would be good, License scanners can properly recognize that information in an easier way.

[maxbachmann]

if I see it correctly the license declaration of GPL-2.0 in LICENSE file (GPL-2.0-only) and COPYING of the final package based on the cpp source is not correct.

Is there a separate LICENSE file version fo GPL-2.0 vs GPL-2.0 and later? As far as I can see the default GPL already includes all the text in regards to "and later".

Another point is that I did not see a copyright notice for David Necas in the distributed package.
But it is an obligation to clearly specify the original copyright holder in the package you distribute in a transparent way e.g. the license file.

2a0f863 hopefully addresses this

[Joerki]

Hi @maxbachmann ,

sorry for the late reply, I missed the notification of your comment. I did not find the authors Matti and Antti in the files. What did I miss?

Regarding the license file:

It contains the GPL-2.0 license text and the appendix "How to apply".

...
                     END OF TERMS AND CONDITIONS

        Appendix: How to Apply These Terms to Your New Programs
...

The appendix is not part of the LICENSE. Instead, the file should have the "Standard License Header" located at the bottom of following page, already on top of the license file (since it is named as "header").

https://spdx.org/licenses/GPL-2.0-or-later.html

The coloring of text on the page helps to show what is part of the license and what is just additional information.

BR

[maxbachmann]

I did not find the authors Matti and Antti in the files. What did I miss?

I added them in the central author list: https://github.com/rapidfuzz/Levenshtein/blob/main/pyproject.toml.
I forgot that there is a historic header in https://github.com/rapidfuzz/Levenshtein/blob/main/src/Levenshtein/Levenshtein-c/_levenshtein.cpp. I think that can probably be dropped though.

Regarding the license file:

I still don't understand how the heck I am supposed to specify it then. So unless someone else opens a PR with how this would have to be done, I can't really do anything about it.