Linux

Yeah, I don’t have good answers for you… I honestly don’t know what the best way to get people into it is. The resources really are not great.

FWIW I think when it does end up clicking everything is a LOT less complicated than it seems at first. Nix is sort of all about building up these attribute sets and then once that really sinks in everything starts to make a lot more sense and you start to realize that there aren’t that many moving parts and there isn’t much magic going on… but getting there is tricky. A lot of people recommend the nix pills, and honestly I think it’s the best way to understand nix itself. If you do earnestly read through them I think there is a good chance you will come out enlightened… they just start so slow and so boringly that it’s tempting to skip ahead and then you’re doomed. They also have a bit of a bad habit of introducing simple examples that don’t work at first which can be confusing, and eventually some of the later stuff seems like “ugh, I thought we already solved this” but it’s building up nicer abstractions. The nix pills give a pretty good overview of best practices in that sense, I think… so maybe it’s the source of truth you’re looking for (or part of it anyway). I think the nix pills are a bit more “how the sausage is made” than is necessary to use nix, but it’s probably the best way to understand what all of these weird mkDerivation functions you keep seeing are actually doing, and having an understanding of the internals of nix makes it a lot easier to understand what’s going on.

For sure! I don’t think we’re actually in disagreement at all, just the limits of text communication :). NixOS is certainly less important to me and I don’t really care if people use it or not at all (it’s nice but there’s enough differences that you have to be aware of that it’d be frustrating to some people — even if ultimately those differences are something that can be worked around… If you’re well versed in nix and Linux NixOS is kind of a no brainer, though). Nix for development (or something like it) is legitimately enough of a game changer to warrant some of the evangelism in my opinion, particularly since as you mention it’s pretty much free to try on any (non-windows) system, and adding nix to a project doesn’t harm non-nix users (more than they’re already harmed anyway, haha). I’ll admit that I worry about how “nix ugly and unintuitive” seems to be a huge problem for adoption, and frankly I don’t blame people for bouncing off of nix (I bounced off of nix in 2011 or so and didn’t come back to it for like 10 years — though it was a bit of a brain worm nagging at me the whole time). That said I think the impression people have of nix being this horrible and completely ugly language (an impression I’ve had in the past as well) is also somewhat untrue. The nix language itself isn’t so bad, but the expectation is for it to just be yaml because “I just want to list dependencies”, which is fair and it might be nice if we had some better abstractions to make that more clear. All of the phases in a nix derivation are confusing and poorly documented, and some operations on attribute sets should probably just have nice special syntax instead of these fancy update fixpoints that the average developer isn’t going to understand… ultimately I’m a little unclear on how much of this is “the nix language sucks and needs to be thrown out” and how much is “we really need a better introduction to what this is and how to use it, especially with some beginner examples and best practices for different languages”. I worry a bit about non-nix nix package managers just from the perspective that it’s really nice to have the one tool to rule all development environments, but maybe fragmentation won’t be a huge problem.