I like that the NixOS packaging system feels like it’s build for Free Software, making source code and Git repositories a first class citizen. You can simply drop a flake.nix into your repository and turn it into a Nix package within a couple of minutes, that’s quite a bit different than the utter headache it is to package something for Debian. Nix packages being free of naming conflicts also makes it very easy to mix and match whatever versions you need, something that’s basically impossible on most other distros unless you resort to containers or virtual machines. NixOS having the largest package collection of any distro is a plus too.
How different (if at all) does Nixos feel as a daily driver, if at all? Is it only about getting used to the system, or does it require to do everything the Nixos way?
Also how does user-level configuration work? Does the upgrade system just ignore your $HOME in terms of version control?
All OS configuration (i.e. your installed packages, services, /etc content) happens in NixOS via a single configuration file in /etc/nixos/configuration.nix. When you do nixos-rebuild switch that file gets read and the OS gets rebuild, which in practice means some downloading and changing some symlinks, so it’s reasonably fast, kind of like GNU stow. The partition table isn’t touched here, that has to be setup manually on the first install like on every other distribution. NixOS will also not rollback the file system if you decide to boot into an older version, everything is done by symlinks and environment variables, so you don’t have to worry about your $HOME disappearing or anything like that.
For daily use that means that you basically never edit anything manually in /etc ever again, except for that one NixOS config file. You also can’t since everything will be symlinks into a read-only /nix/store. For permanent package installation you also need to edit the file and rebuild, you can’t do imperative stuff like apt install …. However with NixOS you don’t need to permanently install anything if you just want to try it, you can just do nix run nixpkgs#emacs or nix shell …and it will run it without installing it (everything goes to the /nix/store/ cache and is garbage collected when no longer needed).
This can become a problem when you are dealing with third party packages that come as self extracting installer files that want to install themselves in /usr or depend on programs being available as /usr/bin/python or whatever, since on NixOS that whole hierachy is empty ( except for /usr/bin/env and /bin/sh). NixOS has buildFHSUserEnv to work around that and provide apps with a normal looking Linux filesystem, but that requires a bit more effort than the usual curl http://…/install.sh | sh hack.
By default $HOME isn’t touched at all and will behave largely like on any other distribution. You can however install packages as user via nix profile install (which behave much the same as apt would, but is local to your $HOME). The other popular alternative is home-manager, this provides basically the same what /etc/nixos/configuration.nix does, but for your $HOME, so you can start systemd services, generate your bash profile with it or install apps locally in your $HOME. Home-manager has to be enabled manually and is probably best ignored until you are familiar with the rest of the system. Home-manager is also less all-or-nothing than NixOS itself, so you can freely chose which dotfiles you want to manage manually and which you want to generate via home-manager.
As for nitpicks when it comes to daily use: Due to everything in NixOS being fully reproducible, NixOS makes little to no use of binary compatibility, meaning if libfoo changes, everything that depends on it has to change too. This requires a bit more downloading than other distributions. After an upgrade, NixOS will also keep two copies of everything around until you garbage collect them. This allows you to just go back to an older version via the boot manager. But it also means that you might need two or three times as much storage as on other distributions, at least until you garbage collect. But generally that’s a worthy trade-off unless you are on an extremely resource constrained system (anything >=32GB storage should be fine).
Finally, if in doubt, install the Nix package manager on whatever distribution you are using right now. You don’t have to go the full NixOS at once, you can install Nix packages on any Linux distribution and play around with it, similar to flatpak.
All package installation and configuration is handled via config files, so it’s easy to keep track of what’s installed. Also makes re-installation convenient and easy (this is also great if you’re fond of unixporn-style setups).
Because it uses config files to manage this, you can also take advantage of VCS.
Instead of having to work with several shitty DSLs to configure your system, now you only have to use one!
Being able to install multiple versions of the same library is nice. With Nix you can just install whatever the fuck you want, really. Want to use DisplayCAL, but can’t because it was dropped from the current release’s repository for still depending on python 2? No problem, just have your flake pull it from a previous release when it was still in the repos; it’ll just work because builds are done in isolation.
The generation system makes updates fearless, since if something breaks you just rollback.
Development is both better and more annoying. There’s no FHS, so you have to set up a dev environment with a flake every time you want to do a project. This is nice because you don’t miss dependencies as everything has to be explicitly laid out in the flake and other Nix/NixOS users can share your flake and get the same exact dev environment (kiss “it works on my machine” excuses goodbye). Annoying both because you’re required to put in more forward planning with your dev environment and also because it breaks language-specific package managers, so you’re limited to supported languages.
You can’t just git clone/make/install stuff from GitHub, as there needs to be a flake first. If the software already has one included, great! If it doesn’t, you’ll be making it. If you need a dependency that currently isn’t packaged for Nix, you’ll be making more than one. If the software is difficult to package, god help you.
Nix documentation can be really lackluster and also assumes you’re Linux-savvy.
Also how does user-level configuration work?
You use a 3rd party tool called home-manager for this. It provides about the same experience as the system config and has more configuration options for software, so should be preferred when it makes sense.
Overall, it’s great if you’re Linux-savvy and is one of the few distros that is legitimately innovative. Said innovation can also be a pain in the ass on occasion though, but still worth it.
Can I use a file in the same place as the nix config to set the configs for i3 for example? Or anything else that would go in ~/.config? It would be amazing to have all the configuration files in one single folder to easily move to new hardware.
Yes, that’s what home-manager is for, your configuration goes into ~/.config/home-manager/ and from that you can generate all the other configuration files that go into your $HOME (either by just copying a read-only version of the raw file or by generating it on the fly from the nix configuration file when a home-manager module is provided)
Definitely Ubuntu, it’s the most user friendly for people coming from other OSes
I’d say Ubuntu again, or maybe Debian. You built up skill and learnt the distro so you want to use something you know for work. On your personal computer, try other ones. I personally picked Arch at this time (around 2012), which helped me “understand” how the OS works, rather than simply use it. I reinstalled it quite a few times and broke the system a lot.
Any distro with a simple package management system. My personal choice goes to crux, but it’s very barebones. NixOS or Gentoo would be fine too The point here is to learn how to build packages by building them yourself, and I feel like the “big” packages managers (apt, yum, dnf) are too complex for that. They also decorelate runtime libraries from headers files, which is a pain to work with as a développer IMO.
But it’s just my personal experience, many new distro popped up since then. Also for reference, I’ve been using Linux for 12 years now, and I run Crux on my desktop, Ubuntu at work, and OpenBSD on my servers.
@Anolutheos@Lolors17 I use Mint Debian edition. I got fed up opening my laptop and having to update when MS said so, so switched to Ubuntu, then Mint, the LMDE and have stayed for 4 years. It's not exciting, cutting edge, etc but neither am I! It just works all the time. Updates are easy and everything is boringly reliable - I love it!
Hopefully LMDE6 is a game changer for the most popular first Linux distro. If the CosmicOS by System76 doesn’t win that title.
My grandparents were 1,5 years with Mint but LMDE5 has now been for 10 months and it is awesome. Literally 0 issues since day 0 whereas Win7 and Win10 caused constant headaches for me over the phone.
Unfortunately I’m not. I’m running numerous Thinkpads until System76 releases their in-house produced Virgo laptop with hot-swappable mechanical keys and open source bios (Coreboot). It’ll also have the trackpoint from Thinkpads.
This is what I recommend for Linux newbs. And they can stay with it if they’re happy with it. It’s also a decently competent Linux distribution which is a hell of a bonus.
I have tried both PopOs and Manjaro. I dont really like both of them.
For Manjaro, I used one of the community edition repacks and many of my peripherals failed to work on boot like Wifi and keyboard. Unfortunately, the week I installed Manjaro was also the week the forums went down so that made things way worse.
PopOs felt quite buggy. I used to postpone every kernel upgrade by at least 2 weeks as something would always break after.
Currently I’m using Linux Mint and everything just works. If I were to start over I would go with Mint, Ubuntu or Fedora. I have found the WM and workflow that I’m comfortable with and would like to just be on a distro that works.
I use Arch. I use the command-line to update, I am very glad that I can do the updates when I do want them. Of course, going over the update list is my responsibility, but such is the power my OS grants me—I can make or break things.
Otherwise, yeah, it’s the customization it offers me. I can make it as janky as I want it to be, or rice it to my heart’s content.
I use arch too. Mainly because of rolling releases. I love the install once last forever philosophy. i also like that arch ships vanilla upstream packages, quickly.
That said arch makes very few choices for you. It aends you to the excellent wiki to make your own choice. So the first install may take a bit of time if you’re new.
To be fair, the fact that Arch makes very few choices for us users is one reason, perhaps the biggest reason, I was hesitant jumping in at the start. A well-meaning friend pushed me off the ledge of hesitation and into the thick of things. Did I feel nervous? Hell yes! But was it worth the frayed nerves? I guess it is.
So many nice things about Arch. I got into Linux with Ubuntu, switched to Debian for many years, and now use Arch.
Why Arch?
AUR provides a huge library of software that natively integrates into your system, including git versions of major components like kernel/mesa so you can test the latest features.
Rolling release means it’s always up-to-date and you don’t have to worry about version-hopping to the next version every release cycle.
Follows upstream projects closely
I installed all my Arch installations with the Calam Arch installer ISO. The one big complaint I see with Arch is the complicated install process, but with Calam installer it’s no different than most other distros.
Now I use Alpine (Edge). I like it because I feel like I am learning more about troubleshooting issues but also because the packages are very up to date.
As a desktop Linux user who doesn’t develop or code in any way, or work with servers, or containers, I found Alpine to be very accessible and the community has bren very patient with my different issues.
Despite how comfortable it is, I think I may end up going back to Debian or finally taking Fedora for a spin. Not for at least a year though.
I think at your age you’re supposed to distro-hop. Catch 'em all. Try another one every two to six weeks. I know several people who did it like this ;-) But your mileage may vary… I like debian.
Well… I just wanted to say you’re supposed to distro-hop when you’re young. I wasn’t trying to imply you should (ever) stop. I think - generally - people are supposed to settle at about 30. Some keep being rebels, though.
OpenSUSE Tumbleweed. I like it for being a rolling release with quality control. On the one hand I don’t like its restrictive defaults but on the other hand I know enough to work with them and that’s given me a leaner system.
I want a boring up to date system with a good KDE desktop that just works (even with an nVidia GPU). Tumbleweed is fine. I don’t want to mess with my computer, I want to use it. I messed with it ages ago when I had to enter xmodelines by hand to make the damn thing work, I’m glad we’re past that.
Same here. Very good KDE Plasma and KDE apps integration, rolling and up to date apps, and very stable at that and if something would go wrong I can easily at boot switch back to a state before the update. Pure gold.
Apart from actual system administration or kernel developing, there’s no real “learn Linux” .
Video/Photo/Vector editing on Linux is not “learning Linux”, it’s learning to use a tool which runs on Linux. You can learn to use Blender, Gimp or Inkscape on Windows. You don’t edit videos/photos/vectors with the Linux kernel. You can even “learn the linux terminal” installing bash on Windows.
You can also install Visual Code or IDLE on Windows and on Linux. Learning to code on Visual Code or IDLE is not really “learning Linux”.
Also going on distro hopping looking for the “perfect distro” many times means the hopper simply doesn’t stick to one long enough to learn how to customize the environment to their liking (which usually means the window manager).
Most of the things you can do on the GUI, even the administration ones are just layers and layers of tools to make things “easier” - and they’ll be different on each distro and release. Command line administration will change much less, or at least less frequently.
Things I consider “learning Linux” are for example:
installing Linux (specially a headless server)
understanding how to use the package managers - again, on the command line
understand how systemd works
(hard core) dive into the kernel workings
understand how grub works
learn the general filesystem structure
learn how to analyze logs
learn user administration and how the permissions (and extended permissions) work
learn how to integrate Linux to a Windows environment (join a workgroup or domain, share storage, authenticate users)
learn how to check resources usage and how to troubleshoot it
undertand the nuances and of partitioning and when they are needed, as well as the different filesystems
etc (and /etc)
And yes, many of those are not strictly “Linux”, but are specific to a Linux system, unlike photo editing.
Mostly good advice. I disagree on the headless server part though. Most people who are interesting in learning "Linux” have a much less reductive idea of what that means than you do, I think. Specifically, I think becoming a comfortable, fluent speaker of a typical Unix/Linux environment and userland is probably the most important thing. I think the best way to start doing that is to just live in Linux, and you’re not going to do that on a headless server. Learning the GUI that your distribution uses to add users isn’t important, but having a GUI where you can run standard browsers and photo editors and such is important, because otherwise, you’ll spend all your time in Windows and never have the chance to develop fluency in all the stuff that is actually important.
Limiting yourself to only using command line stuff I suspect does more harm than good, unless you’re hyper-motivated to learn fast. For most people, the smoother path is probably more gradual. Start with Gnome or whatever and just use the computer. Over many years, you’ll learn a lot of piecemeal things just by becoming frustrated with some problem and learning how to solve it. I do think it’s good advice to do as much from a shell as you can from day one. Instead of using the GUI to copy files, learn to do it from a shell. Just don’t feel like you aren’t allowed to use Firefox to browse the web.
I agree; I mentioned headless server because that would be a more “pure” and general Linux administration - learning how to administer a SUSE Linux using the graphical yast tool won’t translate as well to general Linux admin as if you learn and understand how to fo it in the command line and config files.
And absolutely; one can use Firefox, LibreOffice and any other tool on Linux, but I don’t consider that as “learning” or “knowing” Linux. My wife uses exclusively Linux for 20+ years (because when she left her job where they still had Windows 95, that’s what the desktop at home ran; kubuntu). She does text editing, internet banking, shopping, browsing, printing, everything there (even updates sw through the gui package manager), but she doesn’t “know Linux”.
You can setup a Linux system for a computer illiterate, and they may happily learn how to use it for their social media and streaming consuming, and whatever endusers do in their computers, without ever knowing that’s “Linux”.
Strictly speaking, that already happens. How many Android users know they are running on a Linux kernel?
That’s why when OP said “learn Linux”, I prioritized the admin on command line; as you don’t need to really “learn Linux” to interact with it through automated/graphical admin tools (no shame on doing it, they’re sometimes quicker and more practical than command line).
What I mean is that learning how to use cPanel or Yast is useful, but you’re learning how to administer as system through a tool, which in theory could even be adapted to administer a non-Linux system.
Certainly it’s possible to be a Linux user without learning the things that we would say mean you “know Linux”, but I think the most effective way to learn them also requires being a “user”. Using Firefox on Ubuntu instead of Windows doesn’t teach you Linux, but If you don’t have X11/Wayland and a browser and you can’t do your online banking and social media and Youtube, then you won’t actually learn the “real” stuff, because you’ll spend all of your time in Windows and Linux will feel like homework. Instead, get a full Linux desktop experience that you can do all the things you want to do with, and as you’re doing those things, also seek out opportunities to learn the shell and userland utilities, etc.
I don’t really see Fedora users needing to worry. Fedora is upstream of CentOS Stream and RHEL, so Red Hat will probably love to continue to have Fedora users being their space monkeys / lab rats to find/fix bugs in the OS before pushing to CentOS Stream. Why lose the free labor?
Lab rats is a strong term (not wrong by any means) but people seem to forget that Red Hat is also one of the big players trying to make Desktop Linux better. And when Fedora users report bugs to Red Hat, they fix the bugs not only for themselves, but for the entire Linux Desktop community (they are large contributors to the GNOME Project, as well as making efforts to make Wayland better). Their decisions as a company may be causing community backlash, but without those big players (Canonical, SUSE, Red Hat) Desktop Linux wouldn’t be nearly as good as it is today. I see Fedora as a Debian, but company-backed (say what you want about this statement, but the Fedora desktop experience has been the same for a while, and will not change any time soon.) Fedora is also a Project, not a Product. A distinction that Red Hat takes seriously. Fedora is not profitable to Red Hat (the bug-fixes, as I stated above, benefits everyone in the Linux community, not Red Hat alone), that’s why it’s 100% free (both as in freedom and as in beer). Also, they have full-time employees working on GNOME and Wayland
I started using Fedora after RH killed CentOS, mainly for this reason. However now I feel bit differently about all of this.
At the end of the day, it’s clear RH is not doing this out of good of their heart. They are looking for mutually beneficial relationship, yes. But importantly they are also steering the Linux ecosystem towards that mutually beneficial direction.
And I no longer feel like I can support that. I don’t trust Red Hat as a company to keep innovating and improving the ecosystem in such way it is truly mutually beneficial in long term. I expect that they are mainly interested in directions that benefit RHEL, and allow RHEL to maintain commercially viable, private codebase.
I think that without pushback, they will make desktop linux like so many other Open Source projects: in practice the commercial product is the only really working and well-rounded implementation, because developing alternatives is very complex and requires so much developer time.
So I’d much prefer sending my bug reports to some other community with some other domain. And I’d like to contribute towards pushing the mutually beneficial relationship to a direction where RHEL is just another distribution, and Gnome just another DE. I don’t want a future where it makes sense to say a user is missing Gnome-functionality or RHEL-features, when discussing software that has no reason to be exclusive to either.
If RH is the primary developer of Fedora, and Fedora is the exclusive testbed for desktop-linux, I feel like that’s likely to happen.
linux
Hot
This magazine is from a federated server and may be incomplete. Browse more on the original instance.