[Rd] Why R-project source code is not on Github
simon.urbanek at r-project.org
Mon Aug 25 02:55:51 CEST 2014
On Aug 24, 2014, at 2:22 PM, Spencer Graves <spencer.graves at structuremonitoring.com> wrote:
> On 8/24/2014 10:24 AM, Jeroen Ooms wrote:
>> On Sun, Aug 24, 2014 at 1:10 PM, Gaurav Sehrawat
>> <igauravsehrawat at gmail.com> wrote:
>>> But never mind . Sooner or later.
>> These things take time, but a lot has happened over the past years. By
>> now all activity of r-base  cran  and r-forge  is
>> continuously mirrored on Github, which already gives unprecedented
>> insight in developments. At least several r-core members [4,5,6,7,8]
>> have been spotted on Github, and this years useR2014 website  was
>> developed and hosted completely on Github. It seems like a matter of
>> time until the benefits outweigh the cost of a migration, even to the
>> more conservative stakeholders.
>> However moving development of a medium sized, 20 year old open source
>> project is not trivial. You are dealing with a large commit history
>> and many contributors that all have to overhaul their familiar tools
>> and development practices overnight. There is also the infrastructure
>> of nightly builds and CRAN r-devel package checking that relies on the
>> svn. Moreover moving to Github means changes in communications, for
>> example replacing the current bug tracking system to Github "issues".
>> In addition, several members are skeptical about putting source code
>> in the hands of a for-profit US company, and other legal issues. These
>> are just some of the concerns that would need to be addressed to get
>> everyone on board.
> Am I correct that we could use Git without Github?
No. git is the necessary evil if you deal with Github (almost - Github actually supports direct SVN access as well) - but there is no point in using git alone. I hate git*, but I love Github (despite being very cynical about it before I tried pushing some of my packages there - and seeing the actual impact). The whole point are the collaborative features - and it's hard to get those right (many tried and failed such as GitLab) which is exactly what Github did.
* - git is great for people who like to compile their own Linux kernel every week. Yes, I have done that when I was young - it was fun and I felt great having all the fine control, but once you get a life and just want to get things done, you really don't want to deal with things at such a level anymore. There are better alternatives, but Github placed the bet on git which is what counts in the end. But this is OT so flame me directly if you wish.
> If yes, the planning might involve a cost-benefit comparison of what would be required bring up a not-for-profit alternative to Github (e.g., RGit.org or FreeGit.org or ...), and whether the risks of problems with that would be more or less than the risks associated with "putting source code in the hands of a for-profit US company".
> p.s. Regarding the risks of "putting source code in the hands of a for-profit US company," I would naively expect that it should be easy and cheap for someone to program a server to make daily backup copies of whatever we want from Github. This could provide an insurance policy in case events push the group to leave Github. Many (most?) of those who read this may remember how LibreOffice forked from Open Office. A friend told me that MySQL has a much larger user (and developer?) base than LibreOffice, and every Oracle executive doubtless knows that MySQL could similarly be forked relatively easily.
>> My (limited) experience with these things is that the most critical
>> piece of making such a transition actually happen is not just a
>> general consensus that a is preferable over b, but rather a detailed
>> proposal outlining what the migration would involve, the
>> cost/benefits, a planning, and someone that is willing to take the
>> lead. Only on the basis of such a serious proposal you can have a
>> discussion in which everyone can voice concerns, be assured that
>> his/her interests are secure, and the idea can eventually be put up
>> for a vote.
>>  https://github.com/wch/r-source
>>  https://github.com/cran
>>  https://github.com/rforge
>>  https://github.com/s-u
>>  https://github.com/mmaechler
>>  https://github.com/duncantl
>>  https://github.com/pmur002
>>  https://github.com/dmbates
>>  https://github.com/user2014
>> R-devel at r-project.org mailing list
> R-devel at r-project.org mailing list
More information about the R-devel