I want to find the most sustainable operating system, because computers nowadays waste a lot of energy, because of data collection and data processing. Avoiding unnecessary processes and using resources in a mindful way could reduce the CO2 output on the whole world.
This discussion grew very fast and I put all links to other platforms in the end of the blog article.
This can be done with Fedora.
dnf update —securityapplies security updates only.After that you can cherry pick which applications to update. The cherry picking can be accomplished via an Ansible playbook.
Does it make sense to only update security packages? Or will it even be unstable after updating everything?
I don’t know what Ansible playbook is, but after doing a little research it just looks overpowered. Wouldn’t it be better to have the applications, which must have the fanciest updates in flatpak and than just update flatpak?
btw. could you delete the redundant posts, please. You accidentally post it 4 times.
Yes. “Update for security fixes, and then bump versions only when necessary for features” is how updates are supposed to work, but nobody does this.
RedHat’s release engineering is fantastic. I usually give new Fedora releases a month or two before upgrading my work desktop, but normal updates are uneventful.
Fedora is experimental compared to RHEL, but in the grand scheme of things, it’s a moderate distro. It does more testing then Arch, they try to upstream as much as possible, they don’t ship software with license or patent problems, and it’s a semi-rolling release distro. A few packages are pinned, but most packages get updated as the package maintainer has time, which is usually shortly after release.
That’s up to you. Some people like Flatpak, and some people don’t. I also don’t know how to only install security updates for Flatpak applications.
I use a mixture. Some programs aren’t packaged as a Flatpak, some are only packaged as a Flatpak, and some are better from the distro package.
I’ve run Fedora and RHEL/CentOS for over a decade at this point, and it’s been solid. The times things have gotten weird is when I’ve added 3rd party repos which replace system packages instead of installing into their own path. This problem has mostly been fixed now.
Yeah. I was posting with Remmel, and it’s a little wonky. Four errors, four posts. :\
Isn’t this a hard way? In general I want to update programs, because I hope the bug, which annoys me will be fixed with the next update. So I will do updates for it as long as the bug got fixed, even if I get features I really don’t want. The only way to figure out if an update has a fix is to read all release notes since the version you own. Nobody got time for that.
I totally get it, that if I own a software, which I like how it is, it shouldn’t be changed, but sometimes it’s not possible. I remember that Firefoxs GUI was once pretty lightweight, then they implemented developer tools, which no normal user needs, then a whole customization tool and then a tool to synchronize your data with other devices and so on. And that’s with many software. They reach a point were it is functional, everything work and then it gets screwed up by tons of features only a minority uses. But a browser should be up-to-date, because it could be very dangerous if not.
Do I understand it right that RHEL is like Debian stable, but you have to buy it?
And this is pretty annoying imho, but it might be only the current situation, because I read somewhere that those virtual package managers (I don’t know how to call them otherwise?) will be the future, because there will be only one package to manage, which will work on all Linux distributions. But is this a good thing?
Currently my result for an sustainable experiment would be to use Debian (stable) with AppImage and AppImageUpdate for partial updates. Would you say there is a better solution for a sustainable system? Would you even say Fedora is more sustainable?
Using the minimal viable version is the correct way, but yeah, most people live and die by the
@latestYOLO method.Updates can be done piecemeal in a much more purposeful way to minimize churn, or updates can be blasted out with one command.
You’re correct RHEL is equivalent to Debian stable.
There’s an “up to 16 installs” free tier. I haven’t bothered with it since CentOS is only slightly ahead of RHEL, and I don’t have to figure out entitlements with CentOS.
For a desktop/laptop/workstation, I would stick with Fedora though. It has BTRFS, more desktop software, and more features.
In the past, running RHEL/CentOS as a desktop was a much more advanced project then most people wanted. I was doing lots of custom compilation and upgrade planning for the desktop software I wanted to use. I’m not sure how the new 3yr cadence is going to affect things.
Flatpaks are built for desktop applications. Server applications or development tools don’t really fit into the Flatpak model, and I use server applications and development tools frequently.
It is a good thing. Once a Flatpak is created it is portable across the ecosystem which enhances the software selection for all distros.
Previously, some applications were locked to the big distros, and the smaller distros struggled to port software.
Also, Flatpak is designed to work around some shortfalls of current package managers.
Flatpak can run without root permissions, and it can install applications in the invoking user’s home dir. Most package manager assume the package will be installed on the system, and they don’t have provisions to be run by accounts other then root.
Current package managers aren’t built to version libraries, and this something else Flatpak has addressed.
Debian is fine. I’m just familiar with the challenges of running a point in time distro as a desktop.
I haven’t tried AppImageUpdate. I favor Flatpak over AppImage these days.
Not a good one. :)
It’s as sustainable as any Linux distro. From a user experience point of view, it’s easier to live with on a desktop.
Now that I think about it. A local repo can be setup, and the local repo can be used to update the system.
Mirror the repos to the sdcard, flash drive, or external HD, and then take the drive to each machine for updates. That would reduce the network usage, and reading the local storage is higher bandwidth then the network which would reduce CPU time.
I’m not familiar with
apt, but there might be something similar.It’s interesting that Fedora (and some other distros) are able to do partial updates. This is a huge plus imho, because updates are really energy-intensive (even if not compiled locally) and the reason why I want Debian (stable) is, it gets less updates in terms of, it must be stable before it gets to stable. In the testing branch you would get the bugs as an update and then the fix as another update.
My theory is I can avoid not wanted updates by just using a stable system. And at this point I am afraid that Fedora wouldn’t fit in this logic, because it is like the testing branch. On the other hand it could get the fanciest stuff and I don’t need flatpak/appimage anymore. Okay, I walk in circles again…
CentOS as a workstation? This sounds interesting! I only know it for servers and I thought it was created for servers, but distrowatch says it comes with a DE and it was even discontinued in 2020. I have not noticed anything about it.
And it might make life easier, because you don’t have to learn different things if you have the same system on your workstation and on your server, like different package managers, which reminds me… I still have a v-server running some stuff with CentOS release 6.10 (Final) on it. And the only way it is still on 6 was, that I found it too complex to update and I was afraid I would break it.
But I have to say I’ve never got comfy with yum, maybe because I just used too much apt and pacman in life. Wait… Fedora is using dnf… I thought CentOS is based on Fedora/RHEL. Doesn’t it mean they use the same package manager? Do I have to learn different package managers if I use Fedora as workstation and CentOS for servers?
And I had the problem, that there was no python3 on CentOS and it must be installed by hand, which was a mess and I wouldn’t do it again.
What package manager do you use for servers to have the applications you need? Docker? Is this the way to install stuff like python on CentOS?
In another discussion someone already mentioned, that if you have updated a fedora workstation, it can be used as an update server and it doesn’t sound really hard to configure all other pcs to use the one pc as an update server in the same network. It would definitely deduce the carbon footprint compared to update all workstations over the internet separately. Do you know how this is called, because I forgot it and really would like to get a tutorial for it.
It’s multi-purpose. It mostly gets used for servers, but it can be used as a client.
The CentOS project went through a repositioning in the last couple of years, and things got weird there for a minute.
CentOS 8 is EoL. CentOS Stream 8 is still supported, and there wasn’t significant differences between 8 and Stream 8. CentOS Stream 9 is the latest version, and it’s supported.
CentOS was repositioned to be the upstream of RHEL instead of downstream. In practical terms, CentOS gets packages slightly before RHEL does, and there are more companies and people working on adding software to CentOS then RHEL.
There are a few true downstream rebuilds of RHEL, like Rocky Linux, but it’s too early to tell if they’re going to be around long term.
dnfin included in CentOS Stream 8. There is also ayumcompatibility package installed, which aliasesyumtodnf.dnfandyumwork the same way, as far as users are concerned. Knowing one is basically knowing the other.Going forward,
dnfis the package manager for the Red Hat ecosystem.That’s another thing. Fedora can be upgraded in place. CentOS and RHEL subscribe to the clean install philosophy.
Python3 has been included in the repos since CentOS 7.
It’s a mixture of things depending on what I need.
I do ops and dev work on my desktops/laptops, so there’s Flatpak for GUI tools, GUI and CLI tools from RPMs, services installed from RPMs, some container tools, and custom installs. It’s very much not a basic install.
Servers have stuff from RPMs, containers, and custom installs.
It depends on what you need. If you want to do some Python development for yourself, using a newer version of CentOS and installing Python from the repos is the easiest way.
Containers are a good way to isolate software from the base system, but they add more complexity and systems to manage.
Toolbx is a good way to create disposable environments to work in.
Toolbx
https://containertoolbx.org/
Pkgs.org is a good resource to find packages in the various repos.
https://pkgs.org/search/?q=python3