Project Status Update
Enough with the conceptual posts. Here's the hard status update for 9/3/26:
https://code.foo-projects.org/sofar/ur-project is where the code will live, once pushed. This is the main Rust-based code that contains the producer and the consumer side - In other words it's the builder and the package manager / image builder, all in one. Everything is here. The rest is crumbs compared to it.
why it's not there yet
There's a big difference between "it works" and "it's ready for consumption". I don't want to waste anyone's time with something that isn't ready for consumption, and I find myself having to figure out what it means.
Some hard goals I've set for the initial release:
- there's a live USB image that boots, both on generic x86 hardware and in a VM
- that image can perform a (manual) installation
- that image can perform package updates against a repo server
- if enough disk space is available, you can deploy the ur-project build code and start building
Functionally, most of this is all in place. The devil is in the details. I don't have yet a place to host 310GB (and growing) repos. This means purchasing a costly VPS service, at the minimum. I might have to trim what I publish in the repos. Right now they contain the full artifact chain all the way down from the initial Fedora seed, and most people will not need this. Image hosting similarly has a cost. Images are going to be large. People will want to download them.
Testing wise, everything works, but I've really only tested most things just a few times, and not extensively yet. This will cost time. I need to generate the live image a few times to make sure that it is stable. Streamline the live image experience, maybe provide a good wrapper for the installation so that users aren't stumped.
Package wise, it's got good stuff, and gaps. Xfce4 and GNOME should fully work. Plasma isn't even tested yet, but packaged. There's almost no cloud software, there's no ecosystems like npm/node. No Firefox. Some of this isn't that important right now, so it will have to come later.
These aren't huge hurdles, but, they cost time.
The unknowns
My setups are fairly homogeneous. Most are relatively new x86 systems and I expect no surprises, but, I don't have any systems with e.g. nouveau based gfx. I expect them to work, but do I know? no. There might be some completely reasonable ethernet or wifi cards that just don't work, leaving installs stranded and unable to update. This is likely the biggest risk factor.
Hardening
I'm doing a little hardening to make sure the build system itself doesn't fall over in a normal setup. I had previously all of this deployed on a Fedora system, but two weeks ago I finally switched to hosting the build system on it's own product - proving the ultimate self-hosted gate. But not before I had to quickly package up nginx so that repo hosting over HTTP actually work.
That set off a long list of changes related to FHS shaping the installable builder and permissions, user IDs and all that. That broke a decent amount of designer things enough that now I'm having to revalidate all those paths. It's that part that makes working with packaging efficient. Without that working there's no package flow. Without package flow there's no distro.
Anxiety?
Nah, ship it.
--sofar (aka Auke)