Hey! I’m currently on Fedora Workstation and I’m getting bored. Nothing in particular. I’ve heard about immutable distros and I’m thinking about Fedora Kinoite. The idea is interesting but idk if it’s worth it. CPU and GPU are AMD. Mostly used for gaming.
Immutable distros are all about making thing that were easy into complex, “locked down”, “inflexible”, bullshit to justify jobs and payed tech stacks and a soon to be released property solution.
We had Ansible, containers, ZFS and BTRFS that provided all the required immutability needed already but someone decided that is is time to transform proven development techniques in the hopes of eventually selling some orchestration and/or other proprietary repository / platform / BS like Docker / Kubernetes does.
“Oh but there are truly open-source immutable distros” … true, but this hype is much like Docker and it will invariably and inevitably lead people down a path that will then require some proprietary solution or dependency somewhere that is only required because the “new” technology itself alone doesn’t deliver as others did in the past.
As with CentOS’s fiasco or Docker it doesn’t really matter if there are truly open-source and open ecosystems of immutable distributions because in the end people/companies will pick the proprietary / closed option just because “it’s easier to use” or some other specific thing that will be good on the short term and very bad on the long term. This happened with CentOS vs Debian is currently unfolding with Docker vs LXC/RKT and will happen with Ubuntu vs Debian for all those who moved from CentOS to Ubuntu.
We had good examples of immutable distributions and architectures before as any MIPS router and/or IOT device is usually immutable and there are also reasons why people are moving away from those towards more mutable ARM architectures.
immutable distros don’t provide “increased security” at least not when compared to properly done thing; 2) will lead to a path of even more corporate powered repositories, closed sources and payed ecosystems.
** you should remember that desktop Linux as is, is horribly insecure and any security increase is putting you leagues ahead of most others. immutability is ensuring a minimum level of stability and security.
You are complaining about rambling and didn’t deliver a single argument how immutable distros increase security. Also I think you are wrong big time with 1.
Why do all these immutable distros not support use of secure boot and/or TPM. If there was one that made it a breeze to configure this and made using my AURs easy as well I probably could give immutable a chance. But ATM it all looks like I’ll have to wait until a major corp like Ubuntu made & supported an immutable version so we can get these quirks hashed out.
I’m not sure what you mean exactly but I use Silverblue with secureboot on and a LUKS encrypted drive using a fido2 key. To my knowledge I also could configure the use of TPM to store my key but find that setup not to my liking.
This summary should cover my main concerns with current secure boot implementations on the major distros. Ignore everything else other the linked part. I also would not want to be forced to use grub as the bootloader.
Curious. What did you not like about using TPM to store keys in your setup? I use TPM for secure state validation & automatic decryption of my LUKS drive, it’s great and also acts as a tripwire for secureboot state.
I could build a custom version of Silverblue (u-Blue) to replicate what I already have setup, but none of this would be supported configuration. All this is not entirely to blame on on immutable distros (traditional distros don’t give a damn about secure boot either way), just that to mess around within /etc is a no-no in such a model so to get multiple pre-configured options for secureboot configs/keys that work seamlessly would be a great experience for me.
My (maybe flawed?) thoughts: Why bother with full disk encryption if one could just boot the notebook to undo the encryption?
Using my yubico fido 2 key in combination with a small PIN I can easily decrypt my LUKS drive and know nobody else can decrypt it as long as I have my yubico with me.
My (maybe flawed?) thoughts: Why bother with full disk encryption if one could just boot the notebook to undo the encryption?
If it were that easy to do, we wouldn’t have even bothered with doing disk encryption in the first place. And it’s not like cracking TPMs is a walk in the park.
Using my yubico fido 2 key in combination with a small PIN I can easily decrypt my LUKS drive and know nobody else can decrypt it as long as I have my yubico with me.
This definitely could help in a scenario where an attacker has only your notebook but for it to make a difference your attacker must not be able to access your Yubikey and/or compel you to hand it over.
As long as your LUKS drive is encrypted (TPM or not, Yubikey or not), you are relatively safe from an unauthorized party trying to access your data. Either of these attestation tools add a layer of defense for your encrypted drive.
I believe Universal Blue supports Secure Boot, since they specifically went to make it work for even Nvidia users - I’m assuming it works similarly for the non Nvidia variants or maybe just uses Fedora’s default keys? I’m not too well versed in how SB works.
Then it also comes with Distrobox so you can just spin up an Arch container and use AUR apps through there.
RedHat & Debian family desktop distros use a key that is signed by Microsoft for supporting secure boot. For compatibility reasons mostly as some hardware will brick when the MS signed keys are not found. But I prefer to sign my own keys and enroll them as I currently do with sbctl. I have no need for extra kernel modules/drivers as Nvidia users would (seems like a hacky workaround if the kernel can’t ship the drivers and signing your own kernel makes the situation even more complicated).
However I have been meaning to try Distrobox, if I get around to trying out immutable I will give it a good run.
Ah gotcha, I appreciate the info! I hope that someday a better solution for managing secure boot will work with immutable distros in the future then, so that you have a chance to give it a try (if you want to, of course).
Am already sold on immutable distros as the future of desktop Linux. 8/10 applications that I use today are flatpaks or dockers. That remaining 20% of the work to be done on immutable is what am anxious about.
I jumped. I replaced workstation with silverblue. It feels like installing the early version of the future of linux. From an enduser perspective you do not gain too much going from workstation to silverblue. Yet only if you were already using flatpaks a lot. Installing software into the OS becomes more difficult which is the point. It’s not good for tinkering. For tinkering you should use arch. Reducing the possibilities to fuck up the system sounds great for the end user. I love that you can remove everything but from a business point of view and the responsibility I have when recommending an OS, immutability is great. Moreover it’s more difficult to install snap which is good as well.
There are many good comments here and from what I’ve read immutable seems best suited to the Enterprise IT environment where you don’t want the user fiddling with the system, and you want built in rollback and quick configuration. As well as user data protection.
But for Linux users at home I don’t see any massive advantage. Especially if you’re running a reliable distro like Mint or Debian, or better yet Linux Mint Debian Edition is the best of both worlds.
If you only turn the PC on to watch YouTube, read a document, scan and print, surf the web or game your system should be 100% ok. Unless you’re running Manjaro or Arch.
What I don’t like about the immutable approach is that it turns my PC into a dumb terminal locked by the distro Devs and updated at their will. It’s ok if I have read only on my Android phone because I don’t need to get into root etc. That’s a good place for immutable.
But I don’t want my Linux box at home to be a just an appliance that someone else essentially has control over.
That’s very much an Apple approach. Don’t let the user see or touch anything. They can just be content to change the wallpaper and add a widget. We’ll decide when and how the OS gets updated, what apps they can and cannot run etc.
Ultimately it infringes on user freedom and the very FOSS principles that set Linux apart from the rest.
In short, fine for Enterprise IT but no good for the average Linux user.
What I don’t like about the immutable approach is that it turns my PC into a dumb terminal locked by the distro Devs and updated at their will.
I think you are misunderstanding how immutable distros work. They can be just as configurable as regular distros and in the case of nixOS it is more configurable than popular distros. The point of immutability is to ensure that the system can’t be broken during when it is running by a bad update or install or by user making configuration errors as these are applied during reboot. If the system is broken then a earlier snapshot is booted so you always have a working system. You can setup a regular distro with this atomicity and snapshots but it is not as easy as using immutable distros. Yes tinkering and using native packages is harder in most immutable distros but immutables never were a catch all solution. Use what suits you. I was just a little upset that you claimed that immutables are not in the spirit of FOSS. You can even make your own images(base OS) in distros like fedora silverblue and update your system with those images instead of using what the maintainers provide. It is what uBlue uses
You make a lot of good points, but I have to disagree on the “don’t let the user see or touch anything”. That’s very much not the way immutable distros behave (and I speak mostly about Fedora Silverblue here, I don’t have experience with other immutable systems): you can touch and change anything and often times you have mechanisms put in place by the distro developers to do exactly that. It’s just that the way you make changes is very different from classical distros, that’s all, but you can definitely customize and change whatever you want. I feel the comparison between immutable distros and Apple is really far off: Apple actively prevents users from making changes, while immutable Linux is the opposite – while there may be some technical limitations, the devs try to empower the user as much as possible.
The immutable part (again, only speaking about Silverblue, I don’t know about others) refers to the inability to make changes online (i.e. without rebooting), but you can eventually change whatever file you want. The way it works is you would make your changes in a copy of the current filesystem and at boot simply mount and use the copy. If something goes wrong, you just mount the original at next boot and you have rolled back.
Fantastic. Thanks for explaining that to me. That actually sounds very good and not at all restrictive. Cool. I can see why things are moving in that direction.
If you do a rollback, I assume your data remains? I assume you might need to reinstall apps which were not in the original? Or does it keep apps, data and settings across a restore?
If you do a rollback, I assume your data remains? I assume you might need to reinstall apps which were not in the original? Or does it keep apps, data and settings across a restore?
In CoreOS (Silverblue), /etc, /var and /home (which is in fact a symlink towards /var/home) are regular writable partitions, so your data, configs and personal files are not touched by the upgrade/rollback procedure.
All the packages (and their dependencies) you’ve installed extra are also upgraded/rolledback when you do a system upgrade.
Is AstOS still active? I found this AshOS which is similar but I had never tried any of it. Maybe also inactive don’t really know. Thx for mentioning AstOS, I never thought this is possible on Arch distribution.
No idea! AshOS looks pretty cool too. It would be interesting to have a bigger arch based distro pick up something like this. Other than SteamOS I guess.
Oh nice. I’m going to look into it more. I have invested so much damn time into my current setup and I don’t want to just migrate. So many apps and games are set up the way I want them. Not a distro-hopper here and you get the picture.
I see many people here wondering, why they should consider an immutable system.
As someone, who thought the same a few months ago, and now chose Silverblue, here are reasons why:
Atomic updates: never worry about half applied installations anymore. Either your OS updates successfully, or it will just work like before.
Less bugs and better security: every install is the same, so devs can fix one bug or exploit, recreatable on every system.
Automatic updates (configurable): they get downloaded by the way, without you noticing. And if you reboot anyway, you boot into your updated OS. No waiting times. The system manages itself.
Way harder to break
Changes are easily undoable: if an update breaks anything, you can just select another image and reboot, without recovering anything.
No junk accumulation over time, the OS is kept clean
Clear distinction between “your” stuff and the OS
You can “swap out” the base OS cleanly and keep your stuff. Want KDE? No need to reinstall, just paste one command and delete everything Gnome-related, and you are now on Kinoite.
Flexibility: choose between dozens of different images, like one that replicates SteamOS or Ubuntu, has the MS Surface kernel build in, offers Hyprland, and so on…
And much more!
My #1 reason is, that everything is worry free.
Those advantages above don’t apply to “normal” OSs, even, if I keep everything in Distrobox and Flatpaks.
Immutable OSs aren’t called “The future of Linux” without reason. They usually shouldn’t impair anyone, and make the whole Linux ecosystem better in any aspect.
I’m sorry but none of the above sound different from a regular distro. Maybe I haven’t got the gist. You can have snapshots and atomic updates on a regular distro, you don’t have to reinstall to switch from Gnome to KDE, I can install all kinds of stuff cleanly anyway thanks to package managers, I don’t use root often so the system files are effectively read-only as far as I’m concerned, and so on.
As far as security is concerned I don’t see the big deal, I mean I get why a read-only OS would in theory be harder to break into but it can still be modified for updates so I guess it’s not really “immutable” after all.
What am I missing?
Edit: before anybody points it out, I do know about the rebase layers and I think it’s an interesting approach, but ultimately still gets the same results as packages. It may be helpful for distro builders but doesn’t make much difference as a user.
Yes you can do all this with regular distros but not as conveniently. Especially cleanly switching from gnome to kde and vice versa is a nightmare. And by switching I mean removing one completely(including dependencies) and installing the other.
its an easy: sudo apt install task-kde-desktop; sudo apt purge task-gnome-desktop; sudo apt autopurge
In testing or unstable this can be a problem though.
I feel like, many people just don’t understand exactly how a distro and package managers work. immutable os feels like it allows priotizing only on on a small core part of the distribution which is immutable and slapping everything else on via flatpak or snap.
i don’t like it and i sometimes wonder if we are not going backwards with that approach.
I’m not one hundred percent on the train of immutable, however, i have undertakes nixos and don’t user flatpak/snap. The nix configuration file is where i install everything.
But while.i agree its not super hard to switch DEs on something like ubuntu etc. But one cool thing on nix (which i think you can do on any distro with nix package manager installed) is that you can test the package without installing it at all. The roll bavk id also nice cuz ive had situations where apt gets “broken” ive always been able to fix it with a little searching but its always frightening. Knowing that nix can go back to an old config at anytime makes me a little more comfortable
Funnily enough, I like nix. The concept is way ahead of silverblue and the likes. With nix nothing is hidden behind a compatibility layer. I feel like if we really need immutability, nix is the way to go.
I always depencies left around from the DE that was removed. Maybe it is because my commands are not the right one but I follow what is recommended by the distro wikis. Like if I am using gnome and then download kde just to try it out(without removing gnome), don’t feel like using kde and remove it, I have packages and dependencies leftover from kde when I uninstall it. Neofetch too show an increase in packages even though the only action done was installing kde and uninstalling it
You’re correct. But, and here’s the big but, the whole immutability-thing isn’t something the user should be worried about at all.
On Android for example, the system is read-only too, and pretty much nobody cares too, because it was always designed this way and it doesn’t inhibit functionality.
It is mainly a big pro for developers in how I see it. See, every installation creates some package drift. One dependency here, one extra program there, no problem.
But in sum, there will accumulate hundreds of “bloat”-packages over the years, which add many unknown vulnerabilities and bugs that are completely individual to your setup.
And then it will begin: a program crashes here, there’s your black screen, and every dev on the issue report says " closed, can’t replicate". And after an OS-reinstall, it works again.
And if you want to install KDE on Pop!OS for example, it is highly individual and there are still some packages you didn’t see, and it will be very buggy. Some buttons that are misalligned, misconfigured drivers, and so on.
I tried changing the DE on my normal Fedora one time and even though I thought I did everything correct, I had to reinstall due to screen tearing/ flickering, many misconfigurations, and so on.
On Silverblue, it’s a process of 5 minutes max, and then my setup will be the same as the one from thousand other people.
Ah but on Android they have very rigid rules about partition size, and lots of specialized partitions.
Speaking of which, do you happen to know how immutability is achieved on these Linux distros? Do they mark the system partition read-only, or do they use cgroups, or is it an intrinsic property of the layers?
Package confusions like you describe are always the mark of a poorly designed package system. deb and rpm are positively ancient. deb distros are notorious for multi-repo hell because each repo only has its own limited dependency scope.
You should not have issues like you described on any sane distro. A package is either in a meta package or not. Dependencies should be clear and if something was not explicitly installed it should be cleared out when the thing that depended on it was uninstalled.
Eh, I don’t do anything illicit on the internet neither work at NASA or any other high-security-related job… so I’m in the “Lol” side of this whole story.
This is why fedora had a little bar after rebooting when I updated right? What am I a Windows user?!? This is the extent of my understanding of immutable distros and I am furious with them.
The progress bar on Gnome-based distros like Fedora and Ubuntu was their offline install.
This increases the likelihood of a successful update without borking your system.
You can always deactivate that or update via terminal.
It has nothing to do with immutable OSs. Actually, most of them even update without you noticing, which is quite convenient imo!
I was mostly joking and I might have been mis-attributing the delay. From the time’s I’ve had Fedora, including with KDE, if I update I have a pause during the next boot where I have to let the install finish before getting back to functional. My belief was that this was because the immutable system could not be running while updating, compared to non-immutable where a standard reboot works with a new kernel et al.
Tinkering on silverblue is similar to tinkering on fedora (at least in my experience) just more restrictive in that the read only parts can’t be changed(obviously) and tinkering with packages requires reboots and layering. The good thing is you can rollback to easily undo shit.
Edit: Tumbleweed is not immutable, you learn something new every day, especially from your mistakes 🙃 (it’s still a really nice distro)
Personally really happy with my choice of Immutable Distro: OpenSuse Tumbleweed. To me, who is half a year into using linux, its very convenient to use an immutable system as IF i were to do a wrong command or whatever its super easy to rollback the system (at least on Suse as it uses btrfs-filesystem). Another thing worth mentioning which is also why I chose to go with immutable is that it really teaches you “the good standards” of where to tinker with files and where not to, at least for a beginner like myself this is very nice.
Oh wow, won’t you look at that! 😅 Well that jsut shows my lack of experience I guess. I swear I heard it somewhere and just believed it was. Or maybe I misread and read that MicroOS and Aeon was, therefore assumed Tumbleweed was… My bad!
As you already noted Tumbleweed isn’t immutable, but it is generally delightful! It’s the one I’ve always been most comfortable with in terms of Rolling Releases
I think immutable OSes serve two purposes: For the developer who needs to operate multiple environments at the same time, and for the utter novice who could screw something up otherwise.
This audience, us, is the exactly middle ground. We like tinkering. We like setting things up.
Yeah man I don’t know. I used to think I like tinkering(used endeavour for a few months) but I am enjoying the no maintaince life with uBlue very much. Most of the time the system updates on its own and I am not even aware that the system updated. Same with flatpaks which also auto update so they are always on the latest version provided by flathub when I use them. But I also like gnome so maybe I am not the tinker lover I thought I was
Not true in my opinion.
You can still tweak the image to your liking, you just have to approach it differently.
One of the many things image based OSs offer is peace of mind.
It’s just great to know my PC will work just as fine tomorrow as it did today, and I don’t have to fix anything.
I can definitely see what you’re saying. But if you start to add packages, what do you gain in terms of known stability? Seems to me you might as well then just “be good” about not adding too many packages to a malleable distribution.
If you use the workflow of an immutable system on a traditional one, you have almost all the disadvantages of the first and pretty much no advantages of the second.
The “immutability” (you can still change stuff) is the wrong thing to look at.
I prefer the term “image based”, that fits better. Everytime you update your image system for example, it gets “pulled” or compared to the original image.
On a traditional one, you have your original image you once downloaded, and that’s it.
Over time, it will still change due to updates and stuff. An immutable is basically a “fresh install” every time.
Most immutables use layering, so you still pull the original image, but after that some stuff gets changed.
It is generally strongly discuraged to install stuff (like GIMP and so on) directly. It should only be the last option, like for drivers.
But even when you directly install, you don’t use all the benefits. The OS is still rebuilt every time and package drift barely happens.
And, back to the beginning, the pros and cons.
It’s like with PDFs. Yeah, it sucks that you can’t edit them. But that’s what they’re built for. They can’t be tampered with, but therefore they look the same on every device and you don’t have to worry about fonts, formatting or symbols.
And on immutables it’s the same: some stuff is a bit more different/ complicated for some, but at the same time, they’re less buggy, more secure, offer instant rollbacks, can be customized and rolled out super quickly, and much more. Read my other comment for more information, including customization by building custom images :)
Thing is: on the “surface” it’s not that much different than the “normal” Fedora and it’s spins.
So, if you want something hugely different on the base, I’d recommend NixOS instead. Nix feels like “the new Arch” for me and is the tinkerer’s dream. It appears to be very complicated too, so it should keep you “not bored” as you said.
I personally wouldn’t use NixOS though, as I am just a “casual” user and don’t want to over-complicate everything.
I personally am very happy with Silverblue, especially due to one reason: the ability to rebase to many many images.
As other commenters have stated, there’s a project called uBlue.
It allows you to swap out the base OS (everything except “your stuff”) with one command, so you can rebase to many different community spins and different desktops cleanly.
The uBlue base OS is just Vanilla SB with some QOL stuff added, like codecs and other stuff. It is really a “just works” distro, that manages itself and functions in the background without you noticing.
The other spins give you different DEs, preconfigured drivers, opinionated approaches to different DEs, a SteamOS clone, and so on…
If you want to try Nix, go for it!
Feel free to update us all :).
When I said Silverblue, I actually meant “atomic Fedora variants”, which include uBlue and Kinoite. You can always switch between those with one command and 2 minutes of download time :)
Well, actually this is not the first time me thinking about NixOS. But I tried reading their docs again and… I CANNOT be asked to deal with this. I’d probably be more likely to do LFS than learn NixOS lol I feel stupid now, saying I’ll try NixOS. As much as I want to, the docs are horrendous
tbf the docs are in the format of manuals, i.e. only useful if you already know what you’re looking for or have lots of time. If you don’t, read blog posts and nixos.wiki.
Add comment