I've been working recently on providing nvm integration in Docker base images and on our CI/CD, with a target to make it work as easy for developers as possible.
I think, I nailed it, but... I didn't saw anyone else doing it like that and I wonder now if it might contain any non obvious flaws?
We’re all divided with recent decision to focus on CentOS Stream, which essentially means that stable, professional distro will turn into rolling release now. Also CentOS board members don’t gave us more confidence for the future.
I don’t want to be totally sceptic, I would like to test it on my own and only then, decide if it’s stable enough. But I work mostly with Docker containers and there are no official Docker images with Stream variant. I decided to create it on my own, based on official instruction.
I think a big differentiating factor for people will depend on their patching strategy. For those that pull images and updates directly from the internet, I would expect some surprises and feature-creep. For those that rsync all the repos into their datacenter and maintain point-in-time snapshots of OS repos, their own Artifacts, etc... then use their CI/CD pipelines to validate everything from OS to services, they should have more controlled suprises at build and verify time. Those folks should not see surprises in their production environment in my opinion.
Years ago I switched from Gentoo to Debian/Ubuntu, because I was looking for more stability. I was just tired of resolving problems, that I don't care. This CentOS switch feels for me like exactly opposite direction.
On the other side, for containers where you can just refer to previous version/build, it doesn't necessarily have to be that disastrous. A lot of apps that I support, are just single package JAR files dropped into Java base image. It's quite hard to break it by OS updates.
The worse thing I see here is that they changed EOL time for already released version. We were building some assumptions about it and they're all gone now.
