License is only stated in DESCRIPTION

[jbedo]

The license this software is released under is unclear as there is no license file nor discussion in the README. The DESCRIPTION state GPL-3 but a LICENSE file would clarify things.

[HenrikBengtsson]

FYI, it's a well established and agreed upon standard that it is sufficient to specify the license in the DESCRIPTION of an R package. It's only for some licenses, such as MIT, that you need to accompany it with a LICENSE file.

The License: field in the DESCRIPTION file is validated by R CMD check --as-cran, so all packages on CRAN are known to have a FOSS license (without typos or ambiguities) accepted by CRAN. Bioconductor also have requirements on FOSS licenses, but AFAIK, they unfortunately do not run R CMD check with --as-cran license validations.

Also, CRAN and Bioconductor present the license online.

[jbedo]

Thanks I know this, but this issue is on the github repository and github itself looks in LICENSE to set metadata for the repository, so it's still useful to provide it.

[HenrikBengtsson]

Ok. Then I think one needs to hide from R, i.e. exclude it using .Rbuildignore, otherwise R CMD check will complain, even without --as-cran. OTH, I think the de facto standard is not to do this. Regarding GitHub, I have a vague memory of a discussion to get GitHub to recognize and parse DESCRIPTION files for the license, but I don't know if that was ever done. FWIW, they do recognize inst/CITATION, so maybe there's hope.

[jbedo]

Looks like that's the solution tidyr chose:

tidyverse/tidyr@1e051d3

[HenrikBengtsson]

Two things there: The LICENSE and the LICENSE.md file. The LICENSE file is required by the MIT license, and R CMD check makes it's there. GitHub's 'View license' link points to the LICENSE file, and ignores the LICENSE.md. But I agree, when looking at the files, the LICENSE.md is easy to spot. BTW, my guess is that R CMD check does not allow a separate LICENSE file because of redundancy mistakes, e.g. you might end up with a formal License: field that says one thing and a LICENSE file that says something else.

[jbedo]

Thanks for pointing out the subtleties, though I believe GitHub does link to LICENSE.md if it exists (with preference to LICENSE if both exist). I think the policy of relying on the license field makes a lot of sense from a CRAN hosted presentation of a package, but otoh if landing on a repo hosted on GitHub after a google search (how I ended up here) then it's not immediately obvious. It's a shame both standards can't be more easily supported, either by GitHub supporting the license field or by R CMD check actually reading the LICENSE file and verifying the licence field is correctly set.