IMPORTANT: Relicensing PyCBC

[GarethCabournDavies]

Due to some issues with ESA license compatibility, we are looking at re-licensing PyCBC to use a new license.

To be compatible with ESA software and the EPL licenses that they use, I propose that we use the MIT license, but here are some others' licenses:

• Numpy, scipy, astropy, h5py have BSD-3-clause
• bilby, nessai, dynesty, emcee, Mako, APC's global fit have MIT
• pegasus, python, GLASS, LISAanalysistools are under apache

Minor complications:

• lalsuite and FFTW are both under GPL, so we may need to work out how to proceed without those for the LISA context
• to relicense we need to get agreement from those with the ability to reassign copyright for every contribution ever - this may be hard to do!

Over the next few months I will draft a planning document on the plan moving forward

[josh-willis]

While I appreciate that you are drafting a plan (and that this issue is not that plan), I think more explanation is needed even in this issue about why this change is being considered. Why should ESA be an overriding concern, for example, and not any other funding agencies, especially those that have funded substantially the existing development of PyCBC?

Regardless, as I mentioned when this came up in Slack a few weeks ago, in many cases, and commonly in the US, it will not be the individual who wrote the code who owns the copyright to their contribution, and only the copyright owner can consent to a relicensing. For example, any contribution I made on or after 1 September 2017 has copyright held by the California Institute of Technology.

For anyone who made a contribution while an employee of LIGO Laboratory, you should also look at the LIGO Laboratory Open Source Software License Policy. You may or may not be able to get one of the "exceptions granted by the LIGO Lab Directorate" to cover such contributions.