Sometimes missing license information on anaconda.org

[peterjc]

Consider these examples (mostly tools I am familiar with, picked to show three different licenses):

• https://bioconda.github.io/recipes/khmer/README.html
  • License: BSD-3-Clause
• https://bioconda.github.io/recipes/trackhub/README.html
  • License: BSD 3-clause
• https://bioconda.github.io/recipes/toolshed/README.html
  • License: MIT License
• https://bioconda.github.io/recipes/pyani/README.html
  • License: MIT License
• https://bioconda.github.io/recipes/biopython/README.html
  • License: Biopython License Agreement
• ...

vs

• https://anaconda.org/bioconda/khmer
  • License: BSD-3-Clause (worked!)
• https://anaconda.org/bioconda/trackhub
  • License: BSD 3-clause (worked! Note space not dash as above)
• https://anaconda.org/bioconda/toolshed
  • License: MIT License (worked!)
• https://anaconda.org/bioconda/pyani
  • License: Unspecified (should be MIT)
• https://anaconda.org/bioconda/biopython
  • License: Unspecified (should be Biopython License Agreement)
• ...

Bioconda contains a mix of both "MIT License" and "MIT" (and variants thereof), so initially I wondered if Anaconda.org using a restricted vocabulary - but that alone does not explain the toolshed vs pyani example (which differ in their license_family entry).

Likewise Bioconda contains a mix of both "BSD-3-clause", "BSD 3-clause", "3-Clause BSD", etc - and at least some of the time Anaconda shows them.

Is there a convention used on Anaconda.org about the license metadata which BioConda is not consistently following?

[daler]

This is a good point and something that we've not been good about. #1581 was a start at looking at consistency, and once we have a controlled vocabulary we can add it to the linting system to catch issues with new recipes.

Like you mentioned, anaconda.org seems to be showing "unspecified" if a license_family or license_file key is included. This seems like an anaconda.org bug.

I'm not aware of any license list that anaconda.org uses. Though I guess when there's a question we could use the "SPDX short identifier" from OSI, so for example BSD 3-clause would have dashes as in https://opensource.org/licenses/BSD-3-Clause. Not all licences have those identifiers though.

[peterjc]

I agree whatever is breaking seems to be an anaconda.org bug.

I'm not sure if license_file is important, consider:

Consider these examples:

• https://bioconda.github.io/recipes/wub/README.html showing: MPL-2.0
  https://anaconda.org/bioconda/wub showing License: Unspecified
• https://bioconda.github.io/recipes/nanonet/README.html showing: MPL-2.0
  https://anaconda.org/bioconda/nanonet showing License: Unspecified

Assuming both recipes match the current files on bioconda.github.io and anaconda.org then only one of these uses license_file, but neither has its license shown properly.

about:
  home: https://github.com/nanoporetech/wub
  license: MPL-2.0
  license_file: LICENSE.md
  summary: 'Tools and software library developed by the ONT Applications group'

vs

about:
  home: https://github.com/nanoporetech/nanonet
  license: MPL-2.0
  summary: Nanonet provides recurrent neural network basecalling for Oxford Nanopore MinION data.

[peterjc]

https://github.com/Anaconda-Platform/support/issues/105#issuecomment-298687862

Apparently this should be fixed on anaconda-client 1.6.3.

Does this mean we need to do a push from the Bioconda side?

[bgruening]

Interesting! I don't understand why but yes it seems so. So we need to wait until a rebuild of a package triggers a new push :(

[bgruening]

@peterjc is this still an issue? Please note that the bot is not also checking for the presence of a license file. We have a long way to go to clean license information, but hopefully, our collaboration with Debian and bio.tools will help here.

[peterjc]

All the examples above work now, except:

https://bioconda.github.io/recipes/nanonet/README.html showing: MPL-2.0
https://anaconda.org/bioconda/nanonet showing License: Unspecified
https://github.com/bioconda/bioconda-recipes/blob/master/recipes/nanonet/meta.yaml showing: MPL-2.0

However, as the package has not been updated for over 2 years, that fits. Hopefully the bot will flag this repository shortly as it lacks a license file?

[druvus]

It seems that ONT completely removed the sources for nanonet.

[peterjc]

We can close this then I think