Ambiguous advice on licenses

[ashwinvis]

Here are two points I found to be not completely true:

Slide: License: GNU GPL or GNU Affero GPL

If the cake is a part of the menu, the famous restaurant has to share the recipes of the entire menu.

If I am not mistaken, according to this analogy recipes are source-codes, and cakes are binary packages.

One thing is for certain, this confuses the reader whether the binary package is distributing a modified form of the GPL licensed code or simply linking against it using the publicly documented API. The former would necessitate the release of the source, for sure. The latter is a bit more nuanced.

From xtensor-stack/xtensor-fftw#36 and conda-forge/conda-forge.github.io#1065 this idea gave energy to the common (mis)concept(ion?) that GPL is viral and which in turn means a binary package which links against GPL licensed FFTW has to be GPL licensed.

Arguments for GPL re-licensing

• Mention GPL in the README JuliaMath/FFTW.jl#41 (comment)
• The conda built binaries may need to be GPL licensed, instead of the license of SHTOOLS conda-forge/pyshtools-feedstock#3

Arguments against GPL re-licensing

• This package (cf binary) likley has to be GPL licensed due to linking against FFTW conda-forge/openmm-feedstock#24 (comment) (see also GNU GPL FAQ in this comment)

Personally, now I believe GPL is not viral and does not infect the rest of the code base. Nevertheless, this was a useful lesson to me, and it introduced to me the world of licenses.

Slide: Who can decide about or change a license?

The copyright holder

That is only true if a thing like a separate "Contributor License Agreement" is signed which is often looked down upon in FOSS circles and in any case used only in big projects like CPython. In other scenarios it up to the copyright holder provided they secure express consent from all the contributors.

[bast]

Thank you for the clarifications!

[bast]

I fixed the second issue but am still unsure what to do with the first good point.