Background > Technical definitions: Maybe mention r-universe as a repository

[fkohrt]

I reckon R users a increasingly likely to come across this: https://r-universe.dev/

[NeuroShepherd]

I like the idea of r-universe and including this in principle, but I think something that people will unknowingly run into when using r-universe is that there's no management of package dependencies by r-universe itself. CRAN (and I think BioC) nominally run regression tests when packages are updated so that the dependencies and reverse dependencies of all packages are compatible.

(This isn't strictly true though--CRAN will notify maintainers of reverse dependent packages when breaking changes occur, and will eventually remove packages if they don't receive appropriate updates.)

[NeuroShepherd]

Also the packages uploaded to r-universe don't even have to pass R CMD Checks:

12.3 Are packages on R-universe required to pass CMD check or meet other criteria?
No. R-universe is an open publishing system. The system just builds and deploys R packages from git into personal cran-like repositories. The owner of the universe is responsible for their own policies and quality control.

[NeuroShepherd]

Ok I think I've settled on mentioning it as an in-development project that essentially unifies packages on CRAN, BioC, and user-uploaded GitHub packages into a single repo.

[fkohrt]

👍 Technically, r-universe hosts one repository per user and allows any Git remote, not only GitHub.