earlier replies
Replying to @Benjaming_G
Obligatory ad: If you happen to be on Ubuntu, consider #r2u for your #rstats needs. All the packages from your screenshot (and all their dependencies, for a total of 118 deb packages) install in 25 (!!) seconds. #rstats #automate_and_accelerate
2
6
11
Do you think Rocker will use r2u at some point? Or is it already?
1
1
It's coming. I owe @cboettig @noamross @eitsupi a proposal, and am late. The main question is how to do it cleanly. We're already scattered with r-base vs r-ver, now this splits r-base between w/wo r2u. But it's the way to go. And we'll do it. #rstats #rocker #r2u
1
1
4
Hi @eddelbuettel I hadn't time yet to say how wonderful #r2u is. Now that April arrives, It would be nice to get a frozen version of it for the last days of the R 4.2.x before 4.3.0 (thinking about rocker and reproducible workflows).
1
1
4
Go for it. Get some storage, get some bandwidth, and experiment with point-in-time overlay snapshots. I think @uri_sohn has some latent interest too (but is busy getting `groundhig` away from MRAN). I will continue to focus on the apt-based repository.
1
Yes, @eddelbuettel I will look at my university if they provide some space and bandwidth to do that.
1
2
My #r2u got very lucky in that someone just plain wonderful from my college helped with a (small, as we do not keep history) VM, and ample bandwith is of course a given for @IllinoisStat so it reflects well on us. Your chance to build on top with a related service!
1
1
@eddelbuettel Could you provide an approximate value for the space required on a server for r2u ? Also, an idea of current bandwidth used (should be much lower for "frozen distros", most probably).
1
Disk space for hosting is surprisingly tiny: it is currently only 35gb (as we delete packages that are superceeded by a new version). Bandwidth should be tiny compared to what is offered otherwise, this part of campus also hosts a main CTAN (T, not R) mirror.
1
@eddelbuettel OK, thanks. I think I have space on a server for that. Will experiment next week.
1
1
Replying to @PhilGrosjean @rstatstweet @RussellSPierce @Benjaming_G @cboettig @noamross @eitsupi @uri_sohn @IllinoisStat
The main "problem" as I see it is that we dispatch to apt, so we have to construct 'dated' index files for it -- instead of for R. Not insurmountable, but AFAIK not done before at least in the R context.

Mar 8, 2023 · 2:38 PM UTC

1
1
Replying to @eddelbuettel @rstatstweet @RussellSPierce @Benjaming_G @cboettig @noamross @eitsupi @uri_sohn @IllinoisStat
@eddelbuettel I am not fluent enough with apt (I just use it happily everyday :-) but I can imagine there may be a problem of versions. I have already sent the application to reserve a server for a yearly April snapshot of CRAN + r2u at UMONS. Will get an answer soon.
1