Changing the license of FMS to Apache 2.0

[thomas-robinson]

Dear FMS developers,

In order to make the flexible modeling system more available for general use, we will be changing the license from GNU LESSER GENERAL PUBLIC LICENSE Version 3 to the Apache License, Version 2.0 which can be found at https://www.apache.org/licenses/LICENSE-2.0.txt . We will implement this change with the FMS 2025.04 or 2026.01 release (2 release cycles from now) of the FMS code. If there are any issues with this change, please reply to this issue. If you would like to license your own contribution with a compatible license, please put in a pull request with that update. If you would like to remove your contribution to FMS, please let us know so that we can work out a solution without breaking the code base.

@edhartnett @edwardhartnett @aerorahul @climbfuji @nichannah @SamuelTrahanNOAA @DusanJovic-NOAA @dkokron @harshula @mathomp4 @cxfan1997 @rmontuoro @anton-seaice @MinsukJi-NOAA @benjamindburke @DomHeinzeller @japhir @jkbk2004 @bcc2761 @wramstrom @srherbener @danholdaway @ofuhrer @colingladueNOAA @uramirez8707 @underwoo @thomas-robinson @rem1776 @bensonr @Zhi-Liang @mlee03 @gbw-gfdl @wrongkindofdoctor @menzel-gfdl @GFDL-Eric @J-Lentz @ngs333 @nikizadehgfdl @slm7826 @laurenchilutti @ganganoaa @Hallberg-NOAA @mcallic2 @fabienpaulot @marshallward @cmdupuis3 @abrooks1085 @MJHarrison-GFDL @byrdman1982 @lwhorowitz @lharris4 @uramirez @vnaik60 @nova0002 @alex-huth @adcroft

[japhir]

While I do not oppose this change, I am curious to read up on your reasons for changing the license.

The GPL-3 has been my default license for open science / FLOSS software, because I like that it requires the distribution of source code (prohibiting someone from making a binary and not contributing back) and forcing the use of the same license so that contributing back to the root project becomes easier than maintaining your own fork.

[thomas-robinson]

While I do not oppose this change, I am curious to read up on your reasons for changing the license.

The GPL-3 has been my default license for open science / FLOSS software, because I like that it requires the distribution of source code (prohibiting someone from making a binary and not contributing back) and forcing the use of the same license so that contributing back to the root project becomes easier than maintaining your own fork.

This is a fair request and probably should have been in the original message. Here are two reasons that motivate the change:

1. Some of our newer projects such as pyFMS and cFMS are being released with Apache 2.0. We are keeping the license of these new projects consistent with the license of the projects that they are based on (such as PACE). This will align our licenses.
2. The apache license avoids issues introduced by copyleft licences such as the LGPL. There are several entities that will not engage with copyleft code. Our code is for the public and should be open and as available as possible.

[thomas-robinson]

@lharris4 @uramirez @vnaik60 @nova0002 @alex-huth @adcroft Your names weren't linked for some reason in the original post

[japhir]

Regarding 1.: The GPL-3 is compatible with Apache 2.0 (one-way), which means that even a direct copy of a project (but in this case projects that are based on something) released under Apache 2.0 could be licensed under GPL-3 as well.

Which entities will not engage with copyleft code? As far as I am aware that's typically large corporations that do not want to touch copy-left code because then they would need to release their improvements under GPL-3 as well.

I don't mean to harp on about this but I think it's important to spell out reasoning for a license change on an open science project, also for posterity.

[thomas-robinson]

The rest of the GFDL model codes are currently LGPL 3, so using the apache license means that the models can be released with LGPL-3. Not ALL of the GFDL code is changing, just the FMS infrastructure. This code base will still be able to be used in LGPL projects. Also, as you mentioned, the license compatibility is one way. This code base is a low-level infrastructure, so it is a dependency of other code bases. If one of the other code bases wants to change it's license and includes FMS, that would pose an issue.

The change here is for consistency not compatibility. The newer projects and FMS are codes that are managed by the GFDL Modeling Systems Division, so we are moving to make our licenses the same across projects.

Yes, "entities" would include large corporations. Really, it's any person or business who doesn't want to deal with copyleft issues. If copyleft is a barrier to community development and advancing science, then this is a way to remove that barrier.

[marshallward]

Which entities will not engage with copyleft code? As far as I am aware that's typically large corporations that do not want to touch copy-left code because then they would need to release their improvements under GPL-3 as well.

This has been a real problem on our end. We had to pull out of a hackathon because our codebase was LGPL and the organizers would not have been able to pair us with an industry partner.

[climbfuji]

Which entities will not engage with copyleft code? As far as I am aware that's typically large corporations that do not want to touch copy-left code because then they would need to release their improvements under GPL-3 as well.

This has been a real problem on our end. We had to pull out of a hackathon because our codebase was LGPL and the organizers would not have been able to pair us with an industry partner.

Agreed. We also had issues with this. The easiest of course is to go to Creative Commons Zero since that has the least restrictions, but Apache is better than LGPL (as long as you are allowed to do so as an organization; not all Federal organizations can).

[ofuhrer]

So, I don't mean to be the devils advocate here and should probably not chime in, but my understanding is that under U.S. law (17 U.S.C. § 105) all software produced by federal governement is "public domain", which means that any license in this repo will in any case not be enforceable and absolutely anybody is free to reuse, modify, distribute, or relicense it (even under a proprietary license). So while the license in this repo may indicate GFDL's preferred reuse model, it will not be enforceable and this discussion may be somewhat mute. But having said this, a more permissive license such as Apache 2.0 or BSD 3-Clause may be much more appropriate.

[SamuelTrahanNOAA]

This change opens possible legal questions and I suspect none of us are lawyers; I'm certainly not. Some examples. Do you have permission from all copyright holders to change the license? If not, can you ignore them and change the license anyway? Are there any laws or regulations that could come back and bite NOAA later? Can you even give this a copyright, or must it be public domain?

You should have a NOAA lawyer approve this. They were able to approve other license plans in the past for various NOAA-funded projects.

I'd expect that few people if any would be negatively impacted by the license change.

[thomas-robinson]

The FMS license is set to change with the 2025.04 release currently scheduled for September 2025.