I was talking about regular fedora. It's not that you have to reboot, but you don't get to use those updates until you do. The most obvious example is updating the kernel and its modules.
Shareni
joined 2 years ago
Linux almost never needs to reboot after an update
Doesn't it often need a reboot to apply some updates?
I rember reading something along those lines then I was researching why Fedora installs some updates after a reboot. Most
hmmmmmm
- phallic
- penetrates people
- gives out unsolicited bro science to improve your stabbing performance
OpenBlade reunites with his long lost family
That's such a bad name, I only see lixmaballs.
How do you like it, that's one of the earlier forks, right?
It's far better in theory, but in practice it's got some massive issues:
- non-free packages are taboo in the official guix community
- binary support was lacking the last time I used it (firefox didn't have a precompiled bin for example, and that means you need to leave your browser to compile overnight)
- far less packages than nixpkgs even when you account for the non-free repo
- packages are seriously out of date (I tried using it as an additional pm a few months ago, and debian 12 was newer in a lot of cases)
- essentially no support for some programming languages and package managers (node and npm for example)
In it's current state it's really only good for emacs, lisps, and some other languages like haskell.
You're ignoring the difference between using something declaratory and imperatively. Just because it's difficult to get to that one liner, it doesn't change the fact you'll still only use that one command. Git by it's nature requires you to use different commands to achieve different results. Home-manager allows you to both update your packages and delete all of them with the same command, because that command is "sync the state with the source of truth".
It's much simpler because you're using text files to define the expected state, the cli is there only to tell nix to figure out what it needs to do and to get on with it. Meanwhile with git you're manually doing each of the steps until you reach the desired state.
I only need cd ~/dotfiles/nix/ && nix-channel --update && nix flake update && home-manager switch for everyday package management. It's the nix version of apt update upgrade and install.
nix shell and nix run are pretty useful as well, and you'd want home-manager generations to rollback.
The confusion arises because there are 5 different ways to do the same thing, the non-experimental methods shouldn't be used even though they're recommended in the official docs, and you need to get lucky to get the info that you can use home-manager and that one liner.
Now that I think of it, a guix fork would be far more useful than a nix one. You could forgo some of the FOSS extremism, and allow your users to install it without an ethernet cable, and maybe even on the infidel Operating Systems (even though guix is in the official repo for Debian + wsl).
And I bet guile could really use the attention. AFAIK it's mainly developed by one dude, and he made some impressive improvements. Just check out the release speeches on youtube, massive jumps between versions.
Best of all, the GNU people could focus on building a better core, and choose to adopt only some changes, while preserving the purity of their system.
If anyone is willing to learn a little bit of Guile Scheme - look, the language is great, the project isn't contaminated with multiple scripts, project skeleton is much better, the modules are well written, so why not move over there?
The language is great, but the ecosystem is on life support, and I don't see it getting anywhere close to nix soon. I believe it's especially crippled by being Linux only and forcing free software to the point you're not allowed to even mention the non-free repo in the guix irc.
Random Devs and companies aren't going to use it for their projects, and so there far less maintainers to solve issues like having a node version that's not in maintenance for half a year and 4 major versions behind, or having automated npm package conversions.
Realistically it's currently only useful for a few languages with abysmal PMs, most of which are lisps, and like Haskell.
Don't eat the shrooms!