I’m doing a bunch of AI stuff that needs compiling to try various unrelated apps. I’m making a mess of config files and extras. I’ve been using distrobox and conda. How could I do this better? Chroot? Different user logins for extra home directories? Groups? Most of the packages need access to CUDA and localhost. I would like to keep them out of my main home directory.
No worries, UEFI is definitely a “milage may vary” kinda standard.
I’ve personally only used NixOS and not Nix on a different distro by itself so I’m less familiar with that setup. No system is perfect for all use cases, but that’s sorta the point in Unix-land. I personally have been gripped by NixOS and having to go back to Fedora for some of my old servers has been a pain. They use it like a buzzword all the time, but declarative administration is so awesome. It does have a heck of a learning curve though.
Is the Nix learning curve like Arch’s f.u. user-go cry to rsync/CS Masters expectations, or like Gentoo’s tl;dr approach of “our packagers know how to create sane defaults and include key info you need for sound decisions” approach. I never want to deal with another distro that randomly dumps me into an enormous subject to read because they made a change in a dependency that requires me to manually intervene in a system update, or any OS that makes basic FOSS tools like gimp and FreeCAD tedious.
In my experience its mostly sane defaults and a mixed bag in terms of documentation. For anyone else reading this, https://search.nixos.org/options using this to search for all the built in options is usually a good enough starting point for installing something.
Nix does dependencies very differently, since every program and everything it needs are put into their own checksummed directory, then linked into your PATH as requested in your config. So far I’ve never needed to do anything other than
nixos-rebuild --upgrade switchand only needed to reboot for kernel updates.I mostly work in container spaces, so building things from source, or out-of-repo pkgs, while rare, are done in containers with podman. For example, running Automatic1111’s stable diffusion works perfectly for me in a container with an AMD GPU no less. Eventually I’d like to get into flakes, but their still marked experimental so I haven’t looked too much into it.
Overall the learning experience is figuring out the overall structure of the system, then taking advantage of all the super powerful tooling and consistency those tools offer.