Can anyone speak to their experience with the atomic/immutable versions of Fedora (Onyx, Silverblue)? It sounds pretty interesting.
If you are the kind of Linux user that treats your OS like a pet, with lots of TLC and endless tweaking to get things "just right", then Silverblue is a terrible choice. It is way more difficult and time-consuming to "hack" stuff together at the system level.
However, if you prefer to treat your core OS as livestock (something rigidly structured and reproducible automatically), then I highly recommend giving Silverblue a try!
As a software developer, I run a lot of random software/dependencies/etc and I have historically struggled with my Linux system stability degrading over time as I add/remove packages and config. My goal with Silverblue is to have a solid base OS that does not suffer from package-drift over time. On top of Silverblue, I use distrobox to run many of my apps/bins in a virtual container. That container is where I do all my "hacking" to get things to work and I consider the container to be expendable. If I start having issues with it working correctly, I can just blow away that container and rebuild it without effective my core system, other containers, or the flatpak apps.
I've set all updates to be automatic and almost never have to worry about them. I am on the :latest tag and my computer updates between major versions automatically, and I haven't had any breakage yet. I feel pretty confident in the stability of my system and feel like were the disk to corrupt, I could get back up and running with the exact same setup very quickly.
Silverblue works well enough. The biggest drawback is that you need a different development workflow, all container-based, and you have to use Flatpaks for installing Desktop software.
The biggest questions I had while using it were, "Why am I doing this?" and "Is this worth the trade-offs?".
At the end of the day all the benefits I got ended up being theoretical. Is it "cool"? Yea, its really cool. But does it actually solve concrete problems I have when developing software.
In the end, I decided it didn't.
I think its a much different story for using Fedora CoreOS a server. There the benefits are much more straightforward: reliable atomic OS updates, cattle-only servers, infrastructure-as-code deployment by default.
I still probably wouldn't use them in production because they are a niche technology stewarded by IBM, but still, the benefits are more clear.
That being said, one of the main benefits I am hoping to get out of Silverblue on the desktop is relief from the almost-inevitable system degradation I have experienced with other Linux distros where, over time (and especially with major upgrades), more and more things slow down and/or stop working. I am not sure if I am just doing something wrong (besides daily heavy use), but historically all my desktop Linux installations have become almost unusable after a couple years. I am only a few months into using Silverblue, but I am hoping that the way the system image is rebuilt for each update makes it more stable in the long-term....
It just downloads the new image in the background and applies my local tweaks, which only takes about 5 minutes. Then when convenient, I can reboot into it. Any problems and you don't even need to rollback, just boot into the old release.