Linux

NixOS can’t configure desktop environments, such as Gnome and KDE.

The issue here from my point of view is rather that

1. Desktop environments expect to be configured by the user through their graphical interface for that user only (Nix doesn't touch user configuration) and
2. DEs tend to mix configuration with state (especially Plasma).

Both of these make it close to impossible to configure DEs as Nix configures other services and programs.

For clarification for readers, Nix does allow you to install these environments and associated display managers as in making them available to the user.

But in server environments, tools such as Ansible are orders of magnitudes more comprehensive to everyone who understands the Linux basics. NixOS is therefore dominant in neither desktop nor server environments, but it’s a neat academic project.

Having used Ansible superficially for an Arch server, I disagree. The beauty of Nix isn't (only) that what you declare in your configuration is available to the system. It is the guarantee that it's only these declarations and their dependencies are active. E.g. when you remove something from your systemPackages list, it's no longer installed, you don't need to uninstall it explicitly. Applying a configuration to a system will lead to a known state (if not using flakes, this doesn't apply to derivation versions though). The same cannot be said for applying an Ansible playbook to a system because it's rather additive.

Its nature also makes it a good candidate for development as you can enter well-defined environments regarding available libraries. That's why it's also a good contender there.

I failed to find sources for what I’m about to say, but there was a point where the NixOS stable branch… broke. I’m not sure what went down, but I think that a manual merge train messed smth up. If that’s correct, then NixOS is less stable than Debian.

First off, I haven't heard about the issue. I agree this is something that shouldn't happen, however I'm also having a bit of a hard time understanding what this would mean. You can always boot your old generations and wait for upstream nixpkgs to be fixed; plus, if using flakes, you can also go back to a pinned version of nixpkgs on your current channel. At least that's my understanding. Myself I have never had the need.

There’s also the situation where they store the entire package store in Amazon S3 because someone else paid for it. That someone disappeared, and they expect the community to stem the costs now. If they don’t pay up, NixOS stability is once again dead.

The situation isn't great, I agree, however it should be noted that this concerns the binary cache that keeps compiled copies of all revisions of all packages. Even if it was gone tomorrow, one could still use NixOS as a source based distribution, which would of course require the original sources still being online. The thread is not about the nixpkgs repository — it is fully defined by the Github repository — but about the (arguably very important) binary cache. It would not lead to NixOS being unstable.

I feel like the tooling is all over the place. There are many ways to do one thing, and you never know what’s the right thing to do.

I agree on this point, especially stuff like nix-shell vs nix shell doesn't help, and the language feels a bit opaque. But I guess that's the logical conclusion when your configuration is code — the former is usually assumed to be a fixed pattern while actual code leaves the decision how to do something to yourself. I have thought about the issue myself, but I don't know how it would actually be solved. Though if you don't deviate from the very defaults, it feels like a classic configuration (if that makes sense).

Overall a poor experience.

It was the extreme opposite for me, I switched my main machine to NixOS after many years on Arch (I joined the BBS about 15 years ago) even though there were no real issues with Arch for me just because the concept had convinced me so quickly.

I will say that the experience is not perfect, but that's not an euphemism. I'm very happy with how everything works. The approach might have issues with certain applications (as you mentioned desktop environments; another big one is Steam) but I rather consider this shoddy application behavior than system shortcomings.

NixOS can’t configure desktop environments, such as Gnome and KDE.

That's because those DEs are configured via config files in your home and NixOS is only about system management; it does not touch your home (other than creating its directory).

If you want to configure home files, you use home-manager which has some support for configuring those DEs.

NixOS is stable, so NixOS works best in server environments.

NixOS has a "stable" version and a rolling one. The stable one is bi-annual though with no LTS, so it's more of a Fedora than a Debian.

The rolling channel ("nixos-unstable") is where all the actual development happens. It has similar stability to Arch, though we do more QA through our CI.
You cannot get an update where the e.g. bootloader breaks. That just doesn't happen for us because the update must have passed a test where the bootloader gets installed in a VM and then booted before it can reach you.

there was a point where the NixOS stable branch… broke. I’m not sure what went down, but I think that a manual merge train messed smth up.

Stable is pretty damn stable, we usually only backport security fixes and the odd bug fix. If a bug fix introduces new bugs, that could "break" stable in a way but stable "breaking" in a major way is pretty much unheard of. Again, as I mentioned, basic things like networking or filesystems are tested by CI.

You're going to have to be a bit more specific than that.

There’s also the situation where they store the entire package store in Amazon S3 because someone else paid for it. That someone disappeared, and they expect the community to stem the costs now. If they don’t pay up, NixOS stability is once again dead.

This is way overblown. That someone is still paying. They've just said they don't want to continue doing so (which is absolutely understandable) and asked to find an alternative soon. They're not suddenly cutting us off; this is not an immediate threat.

The NixOS foundation has funds set aside to continue even this very expensive way of hosting the cache for at least a year. Even if they were to suddenly cut us off (which they won't), we'd be fine.

The post you linked is a call to action to find solutions to reduce that cost and to find a way to sustainably finance it without relying on a single sponsor.

More critical would be if the CDN in front of the S3 Bucket (which is what actually serves most paths to users) were to disappear but the provider of said CDN is directly sponsoring us they haven't shown any sign of wanting out. People are brainstorming to find possible alternatives to that setup aswell though, just in case.

We're fine. Even in the worst case scenario, we'd be fine. In the migration time frame, there might be hiccups here and there but that's not too concerning or special.

I feel like the tooling is all over the place. There are many ways to do one thing, and you never know what’s the right thing to do.

Yeah, tooling has been in a bit of a weird spot where we're currently in the transition between the old and the new way of doing things. The new way isn't finished yet however an is still unstable.

The "old way" works just fine though and you can ignore all the new stuff if you don't like instability. I myself still use the old way. Sometimes I use the "new way" too because it's got some convenient features.