I am pleased to announce a small gift from @tulilirockz , @hanthor, and Jason DeTiberus. Our new collective waking nightmare:
It’s basically Bluefin built on CentOS instead of Fedora, but with the same user experience. We will be prototyping this image as achillobator, but will eventually be pushed to bluefin:lts if there’s interest, commitment from folks, etc. No promises. But … Chonky:
How to get it
Snag the ISO from the github repo and follow the directions. It’s fresh, we recommend a VM for now!
Don’t you always go on and on about scope creep and then you go make an entirely new image?!?
Guilty! Some good news here though! This will be limited scope, it’s purpose is to test the new stuff. So we won’t be building nvidia, dx in here because we may bring some of these methods back into the Bluefin repo, or build out of here. Who knows, so we’re going to try it. Plus adding those will be much simpler now.
Though still rough around the edges, it’s surprisingly complete. It took this ninja team 24h to make this, and let’s be real, most of that was working on bling.
It took over three years to build Universal Blue. And now with bootc, bootc-image-builder, and good old bash and some github actions, the team was able to build this image from scratch, FROM quay.io/centos-bootc/centos-bootc:stream10. From the ground up. In 24 hours.
Other than my doc work this was done entirely by new team members, on their own. Upstream development is firing on all cylinders, their upstream documentation is better than ever, so we’re taking the opportunity to get those teams some usable feedback! Knowing that we can do this so much more quicker is really nice. And for those of you struggling with custom images, this will get much less complex for you too.
Special Thanks
Thanks to Carl George for diving in from CentOS, we would have gone through some deep rabbit holes without your guidance. Hat tip to Colin Walters!
It’s actually trying to descope us
Many parts of ublue are ugly, ancient by cloud native standards some would say. Most of it is hiding all that stuff from you since it’s awful. 
We were all in on rpm-ostree moving to a cloud native model, it’s the reason we started the project. But now upstream is in full gear it’s becoming clear that the cruft is getting in the way. We could be going way faster. We have to start from scratch, and as a bonus, we don’t need to churn the bluefin repo.
Rebuilding Bluefin from the ground up is the goal, but we don’t know what that will look like so we just built basic Bluefin and that’s it. Then we’ll look at everything holistically, and with new eyes and contributors and 3 years of experience to have a rethink.
Also related: [Proposal] Consolidate Main, Config, HWE, Akmods, and Kernel Cache · Issue #691 · ublue-os/main · GitHub
Developer Driven
Jason DeTiberus, who worked on this, is a first time contributor, friend, and former coworker of mine who helped build Kubernetes and Cluster API. The main reason we have to try this is that “Enterprise Linux” side of the house has some deep skills, the kind of skills we need. We get these folks helping out, and it bodes well for all of our desktops, we’ll get more good stuff. 
That’s why I think a CentOS Stream based workstation rounds out the family. We’ll have our options of kernels (which we can just push as new tags, etc), and the apps are all flatpaks anyway.
As always, feel free to ask questions! And I’ll hand it over to @tulilirockz who will explain what uupd is.
Also note, we’re doing this on CentOS Stream itself, if people want Alma or another downstream we leave that up to you. That’s why I chose an Achillobator for this codename, she’s going to be where the action is, right upstream. Bigger claws.
