Could r-universe replace retired CRAN time machine?

[skranz]

Dear R-universe developers,

first thanks for the great infrastructure you developed and maintain. I just wanted to float an idea (related to #91) after reading that the Microsoft CRAN time machine will retire (https://blog.revolutionanalytics.com/2023/01/mran-time-machine-retired.html).

The retirement of the CRAN time machine is a pity, as having old snapshots of CRAN including binaries can be quite useful, e.g. to keep old code and data supplements from research articles reproducible, or to quickly install a package that was removed from CRAN. I wonder whether you could imagine that the r-universe framework could offer a replacement for the CRAN time machine. Daily snapshots as on the CRAN time machine are probably quite resource-intensive, but already monthly or even yearly snapshots might be helpful.

That being said, I have no clue how much time and effort it would cost to implement such a replacement with the r-universe framework.

• Does it suffice to generate a Github user like "cran-snapshot-2019-01-01" for every snapshot date, write a script that copies the package sources from CRAN or a particular old CRAN time machine snapshot to correspondingly generated Github repositiories of that user and then use the standard r-universe infrastructure to compile them? That might be similar to the scripts, you use to scrap repositiories from CRAN, but perhaps it is considerably more complicated.

• I guess one problem is that binaries of older CRAN package snapshots probably should be available for corresponding older R versions, so probably one would kind of need an extension to r-universe that allows to specify that binaries should be generated for particular R versions... that might be complicated.

• There might be other problems, I don't see.

As said, I just wanted to float the idea...

[arnim]

CC @chainsawriot

[arnim]

https://github.com/chainsawriot/rang @skranz

[eitsupi]

Maybe https://r-multiverse.org/ 's production repo.