Licence question

[stan-dev]

Licence question

[jgabry]

@jan-glx Very sorry for the slow response. Wasn't ignoring this question, it just got lost in the tons of issues and emails! Anyway, the issue with the license fundamentally has to do with the license for the Rcpp package. @bgoodri told me that https://github.com/facebook/prophet (which uses RStan and hence Rcpp) gets around this somehow, but I'm not sure exactly what they do.

[jan-glx]

I have almost no legal education, but as far as I understand, the Rcpp library and the Libraries are only linked against when a user installs my package so there is no need for me to license my code under GPL because of this (I would need to for binary builds and my lincence would need to be GPL compatible, eg. MIT). However, since rstantools downloads that GPL-licensed make_cc source code file from ratanarm I am restricted to publish my package under GPL.
You could lift that restriction from users of you package in three ways:

• re-implement make_cc.R etc. and licence und MIT or BSD
• ask rstanarm authors to licence these files under MIT or BSD
• perform the download and patching of make_cc.R etc. on the installation of your user's packages

BTW, I think you could publish your package under MIT - depending on a GPL R package does not require your package to be GPL (otherwise all R packages would be (since they depend on base).

[jgabry]

Thanks for following up. I'm afraid I don't have much legal education either on this topic, but some of my stan-dev colleagues know a lot more than I do about these issues. @bgoodri any thoughts on this? If it's true that one of the options suggested by @jan-glx would lift the restriction then I'm certainly in favor of doing it.

[bgoodri]

That is not how the FSF interprets it. The C++ files generated by the Stan parser that link with Rcpp (which is GPL'd) are distributed when a package is built and uploaded to CRAN. See http://www.gnu.org/licenses/gpl-faq.html#IfLibraryIsGPL https://www.gnu.org/licenses/gpl-faq.en.html#GPLStaticVsDynamic

…

On Tue, Apr 10, 2018 at 12:03 PM, Jan Gleixner ***@***.***> wrote: I have almost no legal education, but as far as I understand, the Rcpp library and the Libraries are only linked against when a user installs my package so there is no need for me to license my code under GPL because of this (I would need to for binary builds and my lincence would need to be GPL compatible, eg. MIT). However, since rstantools downloads that GPL-licensed make_cc source code file from ratanarm I am restricted to publish my package under GPL. You could lift that restriction from users of you package in three ways: - re-implement make_cc.R etc. and licence und MIT or BSD - ask rstanarm authors to licence these files under MIT or BSD - perform the download and patching of make_cc.R etc. on the installation of your user's packages BTW, I think you could publish your package under MIT - depending on a GPL R package does not require your package to be GPL (otherwise all R packages would be (since they depend on base). — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#17 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADOrqqRUwiQm-hq1fxYdn72ASGgiWQe7ks5tnNe2gaJpZM4RXoMS> .

[jan-glx]

Yes, if the package is distributed as binary this binary package by CRAN it needs to be licensed under GPL in the opinion of the FSF.
And other than the authors of e.g. the rubias, fourierin, treespaceandtokenizers` packages, I probably would not risk myself to be sued by Rcore or Rcpp authors or any of their dependencies.

However, all that does not matter for the distribution of the source version of my package or at least the source code itself. In the current situation I could licence all code of my package under e.g. MIT except for the files added by rstantools which force me to license the combination under GPL.

[jgabry]

@bgoodri, if @jan-glx is right about source versions then is there a way we can avoid having to pull the gpl files from rstanarm and still make rstan_package_skeleton do what we need it to do?

[bgoodri]

I don't see it. If you are just putting Stan and R code on GitHub without generating C++ files that link against Rcpp or making binaries, then the machinery that rstanarm uses to make it work on CRAN is overkill. You should probably copy prophet's build process https://github.com/facebook/prophet In a future release, we could possibly put a function into rstantools that built such a package locally.

…

On Tue, Apr 17, 2018 at 1:11 AM, Jonah Gabry ***@***.***> wrote: @bgoodri <https://github.com/bgoodri>, if @jan-glx <https://github.com/jan-glx> is right about source versions then is there a way we can avoid having to pull the gpl files from rstanarm and still make rstan_package_skeleton do what we need it to do? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#17 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADOrqmv6EZJtdF2S8Wjiglz8HXD2gvALks5tpXmNgaJpZM4RXoMS> .

[andrjohns]

Closing this issue given its age, but feel free to re-open if it was never resolved

[jan-glx]

It seems like this issue was indeed fixed in e44ffad and related commits (when switching from downloading accessory files from stanarm codebase to copying files from /inst.