Bluefin has even less. I consider this near-zero documentation.

What do you feel is missing from the documentation, can you be specific? You’re examples are too generalized to be actionable.

I would say it is the methodology. To distill it a bit more in the context of bazzite and universal blue:

• Focus on automation (we do this via gitops) - everything is driven by git
• Declarative definitions: all the components of the base images (the kernel, base packages, etc. are all defined up front), and then the custom images (bazzite) do the same thing on top of that. That makes it easier for someone else to start with a small thing and “make my own bazzite” either from scratch, off of a base image, or if you want to just FROM bazzite you can start from there.
• Iterate fast: basically be able to change anything in the OS and rebuild on the spot locally as fast as possible.
• Everything is an OCI artifact

Dude, thank you for this. IMO reducing that down to simply “cloud native” is doing a disservice to how absolutely cool that methodology is.

The methodology IS cloud native, we didn’t invent this. 😼 People will update their terminology, we’re not doing anything new, Linux in infrastructure went through this a decade ago. It’s an update in vocabulary because it’s a shift away from the traditional distro model and has more in common with the rest of industry (k8s, docker, etc) than a desktop. The desktop is just the payload.

We know some people will complain but whatever, it’s our job to help people understand the tech and there are proper definitions for this stuff - The whole “immutables” or whatever slang people are making up doesn’t really make sense but we can’t control what people think, we can just do our thing and keep pushing out updates.

RancherOS doesn’t exist anymore, but a difference here is everything on the machine runs on the metal except whatever workload you have. Here’s people who do a way better job explaining it:

• https://www.youtube.com/watch?v=zF7aTCoWoLQ
• https://www.youtube.com/watch?v=T9aDEF_4vr8

Our systems share the same tooling as Fedora CoreOS so this is probably a better example. You can make custom server images – we build on top of that too, similar to Bazzite but for server nerds: https://github.com/ublue-os/ucore - basically if you can script it, you can make an OS image out of it. Here’s bootc upstream where people are hanging out: https://github.com/containers/bootc/discussions

Hope this helps!

Yes, it’s a container like an app container you would deploy on docker or kubernetes.

It starts with a dockerfile with a FROM fedora, the difference is there’s an entire OS in there, with a kernel and everything. Then an action runs podman build on that container every day, which is then shoved into an OCI registry (in this case ghcr.io).

Then instead of each client doing package updates via a package manager it effectively does the equivalent of a podman pull on your laptop, and then stages the update for deployment on the device. Everything is running on the bare metal on the device, the cloud native part is the build process, pipeline, and delivery. Then rinse and repeat for updates.

It’s a bit like rancherOS except using podman.

Hi I’m the guy who posted the report. Your quote is exactly what it is, we use cloud native server tech to make Bazzite. Things like bootc, podman, OCI containers, etc.

all I want to run is a “normal” PC with a Linux distro.

That’s exactly what’s happening!

I don’t want is to have it running, or heavily integrated in some proprietary-ish cloud.

It does, just not ours, Valve runs that part. 😼 I’m happy to answer specific questions if you have any!