License details

[jbusecke]

I think we are good on the "Does the repository contain a plain-text LICENSE file with the contents of an OSI approved software license?" requirement of the JOSS paper, but perhaps we should look over the actual license text. It says 2017 and has some rather generic statements in there.

[NoraLoose]

Does anyone have recommendations for license details of this package?

• Are we happy with the current license (BSD 3-Clause)? I remember @Hallberg-NOAA having concerns about this kind of license, but I can't remember why. Would an MIT license be a better choice?
• As @jbusecke says above, we should also modify the license text.

[iangrooms]

Numpy & Scipy use BSD 3-Clause, so if it's good enough for them then I don't see why it wouldn't be good enough for us. I'm no expert though. I think @Hallberg-NOAA wanted a more restrictive license, maybe GPL.

[rabernat]

IMO we definitely want a "permissive license" like BSD, Apache or MIT, rather than a "copyleft license" like GPL.

[NoraLoose]

Great, let's stick to the BSD 3-Clause license then. I have created a PR (#96) that makes some minor modifications to the license file.

[Hallberg-NOAA]

We have been advised that there are legal problems with government employees contributing to code with licenses that permit commercialization (like BSD), because by law everything we do is always in the public domain within the U.S., and this can be used to invalidate the license on the software as a whole.

For MOM6, we have been advised that LGPL strikes an appropriate balance between being a copyleft license that is generally consistent with the public domain laws that apply to U.S. government work products, while also not requiring the same license to spread to all derived software.

[rabernat]

I'm going to tag @nbren12 of Vulcan / Allen AI institute. They have some interesting license constraints, and I'd love to get their perspective on this. Noah, would LGPL be acceptable for you? Or is it too restrictive?

Related issues:

• Docrep has GPL license xgcm/xgcm#308
• License issue since cftime split Unidata/netcdf4-python#1073

[nbren12]

LGPL is probably compatible with MIT, Apache, etc, for the typical use-cases of xgcm, but I'm not a lawyer. We have used LGPL projects from our MIT-license work in the past.

They have some interesting license constraints

I wouldn't say we have unique constraints...the constraints are in the licenses themselves. We just have lawyers who occasionally check that those constraints are satisfied.

[nbren12]

That said, MIT etc is on our pre-approved list, whereas we typically have to ask permission to use LGPL. It's up to you how to weight that added friction for users versus adding friction for gov't contributors. Not sure if the situation is identical to MOM6 which is primarily developed by GFDL folks. Maybe ask a lawyer?