As the title and the website say, I’m currently making a distro with the goal that you don’t have to compromise between ease of use + safety and bleeding-edge + performance optimizations.
As I will likely be able to publish the 0.0.1 source code to the GitHub link and the ISO to the website sometime next week, I would like people to be aware of my project so I can have some feedback for further improvements when I release the first version. I have not posted this anywhere else yet, as I value the opinion of this community more than Reddit or whatever.
If you have any questions or any criticism, please voice it, I would love to hear it, good or bad, in the end I want to make a proper OS. (I’ll go to bed now so expect replies to take a few hours from posting.)
CachyOS automatically does bootable BTRFS snapshots with every update. I don’t really see any advantage of atomic distros over this.
Can’t speak much about OP’s distro or CachyOS, but I can talk about how Fedora Atomic differs from simply doing btrfs backups before every update.
First, updates are not the only thing that can modify the base system. Installing packages, installing plugins or extensions, manual user modifications, etc, can all end up bricking the system. Fedora Atomic makes the base system immutable so that the only way of making modifications is via rpm-ostree, which ensures that you always have a way to rollback.
Second, one of the issues with traditional update mechanisms is that over time, updates might not do proper clean-up and leave artifacts. This is a harder problem than people realize, because users can have all sorts of permutations of packages and versions installed, and the update mechanism has to account for all of them. Over time, and many updates, a system can accumulate tons of small update errors and finally fall over. I’ve had this issue myself and heard of this issue from multiple sources. Fedora Atomic solves this by ensuring that every update is like a complete re-install.
Third, since every Fedora Atomic user uses the same base system as their distro maintainers, they can be sure that their base system is well tested. Bugs are more reproducible, and thus can be fixed faster. None of those “works on my system, you’re on your own” issues.
Fourth, this one isn’t a benefit to the user but the community. Fedora Atomic (and universal blue) distros are easy to fork. This is why there are so many flavors now. Bluefin, Aurora, Bazzite, Secureblue, Wayblue, etc. Instead of one generic distro with a huge community relying on a few maintainers, like Fedora used to have, now you have smaller specialized communities with their own leaders, and users can choose which community they align with more. And you can even switch communities, since rpm-ostree allows you to “rebase” to a different distro (eg Aurora -> Bazzite).
My problem with all of those systems is they, by definition, must be very opinionated as to what the base system is. A common counter argument is using various types of “layers” (with differing terminology depending on distro or technology used) but the fact that even for “just” Fedora there are so many bases to choose from shows how that is limited in practice, and slight differences in system preferences makes the “one clean base you can then build from” not as flexible as implied.
I love the idea of them, and definitely recommend them for a new user that doesn’t care what is under the hood, but as soon as you start wanting to change core pieces they start getting more and more in your way, and “rolling back to a clean base” is no longer useful.
I think Nix has the closest ideal to what I would like for myself, but I don’t like the language of its build files and the fact that the recommended way to do things is not the way the official docs describe is too off-putting, so I just run back to my Arch setup. 10 minutes of package selection and installing is more than simple enough for me if I ever need to rebuild and I get to choose exactly what stacks I am using.
With power comes responsibility. Flexibility is power, and the easier it is to modify core system components, the easier it is for you to mess them up as well.
I am not a Fedora Atomic maintainer or lead, so this is just my perspective from the outside, but the core ideology behind it is to have a tailored opinionated experience for user. I remember Bluefin’s lead dev saying something like “I want that ‘defaults’ lifestyle”. It’s the experience of having everything just work out of the box, and be maintained for you.
With Fedora Atomic, the user chooses less responsibility, and thus has less power.
On the flip side you have NixOS and Arch. Maximum power and maximum responsibility. If the world starts moving towards a new window compositor (think X11 -> Wayland), you now have to figure out how to do that migration yourself. If there are customizations that you added that aren’t Wayland compatible, you have to figure out how to work them out. It’s a ton of little decisions. Whereas for an opinionated distro, you just trust that the maintainers handled it and just update your system. Which is exactly what happened for me, as a Fedora user. One day I was using X11, and then the next I was using Wayland, no manual configuration on my part.
This is why I recommend atomic distros to beginners. Beginners don’t have the knowledge to handle the responsibility of maintaining a distro. They don’t know why X11 is insecure, what systemd is, how linux permissions work. They need these guard rails. Once they get more familiar with the ecosystem they can choose to install a more customizeable distro.
Very simple reason: you won’t even notice that something broke since the distro will fall back onto the previous version automatically
I’d rather know what’s going on. Silently working around a problem is how you end up with a system no longer set up the way you think it is.
It’'s not working around the issue, it’s just refusing to update. You shouldn’t need to manually rollback a system that doesn’t boot. And you shouldn’t need to manually troubleshoot or fix it either. Just report the failing update and wait for it to be fixed.
The point of most atomic distros is that it’s set up for you. If you want to manually investigate around and work around the issue yourself, then an atomic distro is probably not for you.
Jup, I think that explains my point perfectly, thank you!
I’d add to this that my distro will just spare you the time to unfuck your system. What you do about the fact that the update failed is still up to you to figure out, but at least you start the figuring out process in your damn DE. That in no way is “silently changing how your system works”, that’s literally just saving you time by doing something that you would have to manually do yourself if it was even a possibility.