I’ve been trying to learn more about the project, how it’s run, how the infrastructure works, and even testing one of the images. One thing I know about is wanting official OCI containers from the project. What else is there that you want upstream to take care of instead of you? Obviously the main area where Fedora can’t help is with anything proprietary, but I’m wondering what else there is that could find a home upstream instead of staying with you guys.
Along a similar vein, is there a reason there are some images like for Cinnamon and Vauxite that are not upstream? Is that something you guys want to keep in Universal Blue? Or is it just a matter of needing more time and manpower to push this upstream? I could also imagine there being different expectations for maintaining an image here versus maintaining an official Fedora spin.
Our biggest issue has always been installer support for OCI images. We’ve been communicating with the osbuild team to be able to make a real offline installer … most of that has just been waiting for those features to land, but if you look at the commits over the past few weeks it’s a bunch of stuff we need:
The seperate desktops we started because Fedora didn’t have upstream images for different desktops outside the main ones. Currently rpm-ostree doesn’t have a concept of installing groups, so when we add a new desktop we kinda have to figure out the packages by hand and that got old pretty quickly. I think moving forward what we wanna do is just base off of the OCI images that are in the quay.io/fedora namespace.
I think it’d be neat if Fedora just made a base image for each dnf group of desktops it supports, and then that base image without a desktop is handy for people who want to experiment adding new/niche window managers. Some of them, like the Budgie one were experimental and we’ll just turn them off when 38 is done now that Onyx is done.
I haven’t had a chance to double check these lately, I see some new ones in there. We should probably audit what’s in there vs. what we’re building, but I’m also unsure of what the status is of getting official builds from Fedora.
Sorry the more I read into your question the more I keep digging (heh). I think when Fedora does this it would remove the need for lots of the things we do with ublue:
Thanks for these notes! Good to have a clear view on where Fedora can help.
That sounds like a great idea! I think the closest conversation so this idea has been the topic of removing Firefox from the atomic spins because so many people end up wanting the Flathub Firefox anyway. If folks are willing to drop the browser, I think there may be lots of flexibility. The limitation here seemed to be just needing to find the time or person to do it.
I’m also tracking Image Builder, but I’ve heard some negatives about the experience and how it works. It’s an ongoing project so I’m sure there could be improvements over time.
One area where I wonder if this will still be limited compared to uBlue is in the proprietary code. If this lives under Fedora, I think it’s possible you won’t be able to building something with non-free software. That’s something to discuss.
Another area that may or may not cross into uBlue territory is image distribution. From what I know, it seems like Image Builder is designed for a user to go and build one image for themselves. It’s not like there will be a list of ready-made images outside of official Fedora spins that serve as starting points. I could see them allowing a wider list of custom images to live in the tool in a similar vein to uBlue, but that’s another question to answer.
But yeah, it’s exciting! Part of the reason I want to get involved is because you guys are pushing this area forward. Maybe Image Builder carries this idea into the future. Maybe it’s nice to have two systems. Maybe uBlue keeps Image Builder on its toes. I just want to see this mature sooner rather than later because it seems like there’s so much potential to unlock. I’m bought in.