Ah, our CI has a check for an MSRV of 1.12. I agree that that's ancient and we could probably make it newer, but we should also check if an MSRV bump would break any users. As I mentioned in the other PR, it turns out that there are many of them!

I'm less worried about moving the MSRV forward here, but admit, I can't think of any particularly good way to determine the downstream effect.

My personal observation is that one can't reasonably expect to pull new crate versions off crates.io and have it work with an long past stable release of Rust, so if anything, breaking the long tail of minority users stuck on the ancient rust compiler version (why?) I think at some point it's reasonable to leave behind.

As a project, at some point it becomes unbearable to maintain old releases and requires lots of workarounds (e.g. hack in the 1.12 test that avoids rand 0.8 existence), and that maybe it'd be better if we agreed on some sunset window for the MSRV. 1.12 is 7 years 4 months. If we agreed to a ongoing sunset of 5 years for MSRV that would give us guidance on the issue going forward and allow us to move as far forward as 1.32?

In either case, I'm fine with us bumping this to 1.17 now. Can you update the github workflow for that test so we can get this in?