I’ve observed some notable improvements when benchmarking with the CachyOS kernel on NixOS via Chaotic’s Nyx using moderately old hardware:

https://programming.dev/post/38304031

Haven’t yet tried replicating the same comparison on newer hardware, but would be interested to see what others have tested. Any observations?

Some poignant questions for these new platform requirements:

• How do you anticipate this being used against journalists and advocacy groups?
• What research and statistical quantification will be done to evaluate the amount of harm these restrictions can inflict?
• What precautions or safeguards will users have against malicious state actors or capitulating corporations?
• How can developers protect themselves from liable damages due to service interruptions caused by third party verification?
• Do you foresee legal restrictions in rollout due to national security concerns from differing nation states?

One thing I appreciate about NixOS is the ability to use overlays and override package sources. For example, overlays can be used to selectively install unstable and stable packages alongside each other:

• https://discourse.nixos.org/t/mixing-stable-and-unstable-packages-on-flake-based-nixos-system/50351

While there may be caveats, this approach has been working for me just fine, as I can install VSCode from unstable to get the most recent monthly releases as they roll out, but then pin the rest of my desktop environment to stable to limit anything else shifting underneath me unexpectedly.

If there was a simple Debian based distro that I could declaratively manage via a single config file, I think I’d try it. I.e. not using Puppet or Chef that can only bootstrap a system state, but something to truly manage a system’s entire life cycle, including removing packages and anything littering the system file tree. But since there isn’t, I’m using NixOS instead.

Having a DSL to declare my entire system install, that I can revision control like any other software project, has been convenient for self documenting my setup and changes/fixes over time. Modularizing that config has been great for managing multiple host machines synchronously, so both my laptop and desktop feel the same without extra admin work.

Nixpkgs also bolsters a lot of bleeding edge releases for the majority of FOSS packages I use, which I’m still getting used to. And because of how the packaging works, it’s also trivial to config the packages to build from customer sources or with custom features. E.g. enabling load monitoring for Nvidia GPUs from btop that many distros don’t ship by default.