Thanks! Sounds good. However, we already have some environment variables, see https://github.com/r-lib/remotes#environment-variables

Environment variables are better in some ways, e.g. subprocesses inherit them, and you can set them globally for all R sessions.

Do you have a particular reason to have an option instead of an env var?

Using an R option just seemed the most natural mechanism when I first thought about it, but I'd be entirely open to using environment variables (especially if we want / need to think about subprocesses). Would using e.g. Sys.getenv("REMOTES_UPGRADE", "ask") be better?

Would using e.g. Sys.getenv("REMOTES_UPGRADE", "ask") be better?

Yeah, I like that. Can you please also add it to the list of env vars in the README? Thanks!

My main concern with this is you lose the auto-documenting behavior of c("ask", "always", "never").

An alternative would be to keep the arguments the same and add logic to

https://github.com/jimhester/remotes/blob/8dd72c74ae34e04851f34f5b7b33329572107a43/R/deps.R#L533-L547

We are already changing the default for non-interactive sessions, so maybe just adding it there is ok.

We could have upgrade = c("default", "ask", "always", "never"), and then "default" would do Sys.getenv("R_REMOTES_UPGRADE", if (interactive()) "ask" else "always").

I've updated the PR with the new "default" option -- let me know how it looks.

This LGTM

Thanks!