[Rd] Implications of a Dependency on a GPLed Package
christian at sigg-iten.ch
Fri Jan 25 19:17:11 CET 2013
The GPL FAQ section that I quoted addresses all the points that you raise:
> There are some questions that need to be answered, since these types of questions have to be answered within specific contexts. For example:
> 1. Is your package "pure R" code and contains only code that you have written?
The quoted section of the GLP FAQ concerns purely interpreted R code. Issues of authorship are not part of the argument, but they of course have to be fully resolved.
> 2. Does your package contain code requiring compilation (eg. C, C++, FORTRAN) that links against other GPL libraries, R or otherwise?
> 3. Are you including any other source code, not written by you, that is GPL?
No inclusion of third party source code, only making use of functionality of a GPLed R package.
> If your package is "pure R" (no compiled code and no R code from someone else that is GPL) and contains no linking (in the compiler sense of the term) to R's libraries or the libraries of other GPL licensed code, you are free to distribute your package under any license you wish and even restrict use. Note that I am focusing on the GPL here and not other "looser" open source licenses.
> The same applies if you include a DEPENDS line in your DESCRIPTION file, where other GPL packages are listed. That has no implication for the licensing of your package unless you are linking against libraries in the other GPL packages.
It is my understanding of the quoted GPL FAQ section that the FSF does not agree with this interpretation of the GPL.
> The so-called viral part of the GPL largely comes into play when you link against other GPL code or possibly embed other GPL code within yours and distribute that product. In that case, your code, at least the part that specifically links against other GPL libraries, would have to be licensed under the GPL or a GPL compatible license as well. Albeit, even there, we are talking about distribution, not use. You could develop a package for "internal" use only, in which case, the license used is irrelevant.
I provided examples of packages that are already published on CRAN, which are in violation of the GPL according to the FSF interpretation, provided I understood the GPL FAQ correctly.
> There are non-GPL and non-GPL compatible packages on CRAN and this topic has come up for [heated] discussion in the past. The CRAN maintainers have not placed GPL or GPL compatible only restrictions on the packages on CRAN. That there are such packages on CRAN is not a legal issue vis-a-vis the GPL, but a philosophical one, which is where the heated discussions tend to arise from.
I am aware of those discussions, and am not advocating for a change of the current practices of the CRAN maintainers. But I if the current practices of the R project are in conflict with the FSF interpretation of the GPL, they need to be addressed.
More information about the R-devel