Updated: December 2, 2019
I don't like repeating myself. Or even linking to my own articles. I find the exercise tedious if somewhat necessary to help readers connect the dots. Now here's a theme that I've mentioned so many times it isn't possible to encompass all the relevant references from the past thirteen odd years of Dedoimedo: I want a super-stable desktop with a lifetime support that exceeds hardware life.
Sounds simple. And yet. Throughout my Linux adventures, I've used a tiny number of Linux distros in a serious, production fashion on the desktop. OpenSUSE, Ubuntu, Kubuntu. A little bit of MX Linux lately. But that's pretty much that. The golden formula seems so hard to nail. Distros are either modern and fickle or stable and old. You sort of can't get the sweet spot in between. Or maybe you can. CentOS 8 Stream.
Stream vs non-Stream
Several weeks ago, I tested CentOS 8. It was a pretty solid distro, and the promise of ten years of support is definitely appealing. Top that with some partisan customization, and you can have a very nice desktop, even though this system isn't intended to be used in the classic desktop function. Without extra repos, ordinary users will struggle to find the modern, everyday joy they need. And even when it comes to work, the repo content can get obsolete.
Over the years, CentOS changed, adding more "recent" stuff, like newer versions of the browser and mail client. A decade is a long time to not upgrade the software base, but then it's also extremely risky, especially if you promise no ABI/API breakage over that period. Which means users must choose between bleeding-edge Fedora with its ultra-short life span or the turtle-like CentOS.
The compromise - and a solution - to this quandary could be CentOS Stream. This would be the sort of rolling edition of CentOS, somewhat similar to openSUSE Tumbleweed, which would be a stepping stone between Fedora and CentOS proper. The advantages of this idea are many: people can't or won't commit to Fedora in the work environment, but they could dedicate resources to running a mode "modern" CentOS, because they still have the basic package of longevity, stability and support, with only some changes in the software bundle. That can be controlled and managed and tested. Happy times.
In the Linux space, Ubuntu solves the problem with more frequent LTS releases with shorter if still quite respectable (free) support cycles, and the ability to use additional repo channels that contain newer software versions. Indeed, this model hasn't really been replicated elsewhere, and CentOS Stream represents the only other venue where you get enterprise and desktop features side by side. So we tested.
Same same but different but still same
I performed the installation and setup, much like the regular CentOS 8. There were no differences, including the system customization thereafter. Once I had the desktop configured, I added the EPEL and RPM Fusion repos, I installed the necessary packages to manage extensions, I changed the desktop layout, and installed some extra software.
Then, I looked under the hood, and I saw kernel 4.18. This is the first big divergence from the main release. And the thing is, this early on, the delta between the two versions is quite small. But it will become meaningful later in the distro's life.
But in a way, I'm cheating. Because Stream is intended to help developers and system administrators gain access to fully supported and sanctioned software that exists in the official channels. By using the third-party repos, I've already broken the seal, so to speak. And then, I showed you how to install the mainline kernel in CentOS 7, which kind of "streamified" my distro from the old 3.X kernel family to the new and hot stuff.
These kind of things - desktop software excluded - will be possible in CentOS Stream natively. But then, if you need the ordinary applications for workstation-type activities, you might need the third-party channels anyway. Still, you can blend. Serious components from the Stream repos, and desktopy material from the unsupported extras. However, this is a fragile balance, and given the library dependency mess, this is something that might not work out in the long run. However, for common, desktop use, that's good. In fact, very good.
CentOS 8 Stream looks like a nice, smart project. Whether it's going to grant the intended users, i.e. not desktop folks, the necessary levels of flexibility and stability and modernity, well, time shall tell. But it is a sensible idea, because at the moment, the choice is one between austerity and unpredictability.
I am testing from the desktop angle, so the considerations are definitely different - and the benefits smaller if any. But then, I'm thinking. Could I perhaps commit this to some production or semi-production desktop machine, and see whether it offers long-term value? In general, I don't like the concept of using third-party repositories on systems used for serious work, but I'm tempted enough to see how well this pans out. Hm, we shall see. The end, cliffhanger style.