Dual licensing R packages with code and data
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.
Licenses are an important topic within open source. Without licenses, information or code can be publicly available but not legally available for reuse or redistribution. The open source software community’s most common licenses are the MIT license or the GNU GPLv3.
When you read the MIT or GNU license, you can see they are rather specific:
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”) [MIT License]
and
The GNU General Public License is a free, copyleft license for software and other kinds of works.
They aim to cover primarily software, not other forms of information such as, for example, data.
Licensing only code in R packages
Given the importance of a license for redistribution, CRAN requires R packages to contain an open-source license.1 As one of the most important distributors for R packages historically, this rule has become the de facto standard.
An R package is licensed appropriately when the license is stated in the metadata DESCRIPTION file and, if necessary, in the LICENSE file. We recommend including the LICENSE file, with the license year and exact copyright holder (for example, an organization or a specific individual). Some R package developers choose to include a copy of the license but this is not bundled in the R package when it is built. You can find an example of a DESCRIPTION file here and find an example LICENSE here.
Licensing only data in R packages
It can be the case that an R package is primarily used to bundle and share a data set, for example to allow a user to easily download the data and load it into R. For example, gapminder and palmerpenguins. As an aside, some in the R developer community2 dissuade the use of R packages used primarily for data, often referred to as “data packages”, and instead advocate to host data online and use the several API packages available in R to access the data.
Software licenses do not apply to data. One piece of data cannot be copyrighted, as facts cannot be copyrighted, but a collection of data in a database can come with rights. In Europe, UK, and Russia, databases can have rights if they are a substantial and original piece of work.3 This is called the “sui generis” database right. The USA and Brazil do not recognize this database right. Data compiled from various databases that themselves are copyrighted would need to follow the licenses of those respective databases.
Data need licenses such as the Public Domain Dedication or the Creative Commons Attribution to maximize redistribution. Such general licenses help minimize differences among countries that do and do not recognize database rights. Given that any copyright limitations (even attribution) mean they only apply when the database right is recognized, we recommend the most minimal option: The Public Domain Dedication. This levels the playing field for reuse and redistribution, no matter the jurisdiction.
Licensing code and data in one R package
But what to do if your R package has both code and data as primary objects of (roughly) equal importance?
In these cases a software license inadequately covers the data, and a data license inadequately covers the code. Dual licensing can help resolve this issue. This means there is one license for code (for example, MIT license) and another license for the included data (for example, Public Domain Dedication).
After conducting an online search, dual licensing for R packages seems rare. An interesting example of dual licensing is the igraphdata package, which contains several licenses: One for each dataset included in the package. Similar to igraphdata, in our own epiparameter package we dual licensed the code and data as well. We licensed the code using the DESCRIPTION file and used the LICENSE file to license the data under CC0. Concretely, we include this additional text in LICENSE to clarify the dual license and that we recommend citing the original source regardless:
All data included in the epiparameter R package is licensed under CC0 (https://creativecommons.org/publicdomain/zero/1.0/legalcode.txt). This includes the parameter database (extdata/parameters.json) and data in the data/ folder. Please cite the individual parameter entries in the database when used.
When including data in your R package from other sources it is important to check that the license of your package and the data is compatible4, or that the individual data license is clearly stated, as in igraphdata. For epiparameter, we consider model estimates as facts (that is, not copyrightable).
This blogpost helps explain a pattern of dual licensing rarely seen in the wild, and how to implement it in a CRAN conformant manner. By sharing this, we hope that those cases where people want to license both code and data, have a resource we wish we had while exploring this topic. Whether or not data packages and combined code + data packages should exist, is a question for another day.
Footnotes
For a full list of license accepted by CRAN see: https://svn.r-project.org/R/trunk/share/licenses/license.db and they also accept stating the license as “Unlimited” for unrestricted distribution.↩︎
A discussion of data and R packages can be found here: https://github.com/ropensci/unconf17/issues/61. This thread is used as an example of some thoughts on packaging data in R but we acknowledge it is from 2017 so the opinions of the individuals in this thread may have changed.↩︎
To see the legal definition of the database right in Europe, and what constitutes it, see the European Union Directive 96/9/EC↩︎
Reuse
Citation
@online{lambert2024,
author = {Lambert, Joshua and Hartgerink, Chris},
title = {Dual Licensing {R} Packages with Code and Data},
date = {2024-09-23},
url = {https://epiverse-trace.github.io/posts/data-licensing.html},
langid = {en}
}
R-bloggers.com offers daily e-mail updates about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.