Base on Debian stable instead of Ubuntu LTS

[irondave]

Basing eOS on Debian Stable instead of Ubuntu LTS can offer several distinct advantages, particularly if you’re aiming for a system that emphasizes minimalism, predictability, and a truly community-driven experience:

Unmodified and Pure Base:
Debian Stable provides a relatively unaltered, “pure” environment. Unlike Ubuntu LTS, which layers additional patches and customizations from Canonical, Debian’s base remains closer to upstream. This can simplify maintenance and allow eOS to implement its own vision without working around extra modifications.

Rock-Solid Stability:
Debian Stable is renowned for its rigorous testing and conservative update cycle. This reliability means fewer surprises and a more predictable platform—ideal for a desktop OS where stability is paramount.

Minimal Bloat:
Debian’s default installation tends to be leaner. This minimalism gives eOS developers the freedom to add only what’s needed, ensuring that the final product remains lightweight and performance-optimized, whereas Ubuntu LTS often comes with additional packages and features that might not align with eOS’s streamlined approach.

Enhanced Customizability:
With Debian’s more “bare-bones” nature, eOS can build a customized user experience from the ground up. This level of control allows developers to tailor every aspect of the system, which can be harder to achieve when starting with a base that includes extra, pre-configured elements like Ubuntu LTS.

Community-Driven Ethos:
Debian is entirely community-driven, which can resonate well with users and developers who value transparency and collective input over corporate decision-making. This can also lead to a more flexible and open ecosystem for future enhancements in eOS.

Overall, choosing Debian Stable as the foundation for eOS can result in a leaner, more stable, and highly customizable system that stays true to a community-first philosophy, avoiding the potential bloat and corporate-imposed layers found in Ubuntu LTS.

Thank you.

[teamcons]

Without wanting to sound rude - is this written by chatGPT/an LLM ?

[reece146]

Often thought about what it would take to re-base on Debian. Debian is a better philosophical fit since it is a community driven upstream source.

To kick the can around...

In terms of the effort required, one of the obvious things would be the age of the underlying OS components and if they make a difference to compiling the Elementary specific tools, libraries, apps, etc. Perhaps once an iteration or two of releases went by there would be alignment of all the components and the Debian cadence would be fine.

Another thing is that the Ubuntu upstream HWE system gives Elementary an 'easy button' of sorts keeping the kernel, firmware, etc. up to date independent of making a full release. Debian does not have an easy HWE method and I find the backports system is klunky at best. Having experience of running ZFS enabled Debian hosts in the past with backport kernels, it wasn't matter of if but when a system or two wouldn't boot and needed hands on sysadmin level interaction to correct. Maybe things are better nowadays.

It's a huge undertaking trying to have an Ubuntu and Debian based version at the same time; not just the creation but also the maintenance and support aspects. Linux Mint is doing this currently but I get the sense that is a much larger team.

2¢

[luizmichelin-cmyk]

I won't be technical to discuss this one - but practical instead.
While I can appreciate all the work behind Ubuntu, I'm not really a fan of the system.
Noting so, I can assume that the reason behind building Elementary over Ubuntu resides on the fact that Ubuntu was built to run a larger sample of computers, and opposed to it, setting Debian is, sometimes, a patience exercise (and not even all the patience on Earth ensures Debian will run in all machines).
In that sense, scaling Elementary through the Ubuntu technology seems reasonable, but (as I noted, I'm not an Ubuntu fan), I agree with @irondave concerning the overall stability behind Debian.
Canonical is completely entitled to their commercial decisions, but - in my opinion - some of these decisions are pushing them way too much closer to Microsoft than to Linux ideals.
We all know that, invariably, such move can only fall in trade-offs and, usually, money wins over users.

[o-alquimista]

I would love to have the newer packages from the latest Debian stable. Ubuntu LTS is so old.

[qntnlq]

I would definitely love to see elementary rebase on Debian instead of Ubuntu. Moreover, I feel like Ubuntu is going on a path that's moving (even more) away from elementary, from a philosophical and ethical point of view. e.g. their stance on 'thoughtful' AI compared to the rules at elementary (which I support).
However, it seems to me like a tremendous amount of work for little gain (and I could very well be wildly underestimating the amount of work while overestimating the gains in the same extent - maybe @danirabbit has nightmares just from thinking about such a rebase haha) and as @reece146 stated, maintaining 2 OS versions sound like a huge undertaking which would, I think, spread the dev team too thin...

[reece146]

I was trying to find this discussion yesterday...

Tuxedo recently announced they are moving from Ubuntu to Debian Testing (current: Forky) as their base. I wouldn't be surprised to see other distros take this tact as Ubuntu moves more and more away from what made it attractive. I know the KDE Neon team has publicly talked about the friction using Ubuntu as their base.

From a practical sense, having run Forky for a few months relatively recently, there's nothing very unstable about Forky. It seems like just plain-jane Linux OS infrastructure, especially compared with more bleeding edge distros. Caveats apply ofc.

Still a huge lift.

[qntnlq]

Well, if elementary ever rebase on Debian testing and becomes a semi-rolling release distro, then I'm switching all my machines on it asap haha It would be too good to be true

Though I wonder if this is doable. For TuxedoOS, they stated that they would get ahead of Testing and backport from Sid when patches take "longer than [they] consider acceptable", which will notably happen when Testing freezes each time the new stable is getting ready for launch. Is such a workload acceptable for elementary ? Maybe they could freeze Debian Sid just like Ubuntu does, but then it's additional work too. Maybe they could also just let the freeze happen and wait for the new cycle to begin, but I don't know how this is for security (real question, do security patches reach Testing during the freeze period ?)

[Quartz-Money]

I also strongly agree with this idea; as other users have previously noted, Ubuntu is drifting further and further away from the elementary philosophy. Switching the base directly to Debian would allow that philosophy to be—shall we say—better upheld. That said, we cannot overlook the fact that the Ubuntu base—despite occasional missteps caused by poor decisions from Canonical—also aids elementary's development by not being as "bare-bones" as Debian.

[qntnlq]

Debian is currently voting a resolution about LLM usage. Granted the proposal A or C get accepted (which I hope; if B is selected I'll have to find another distro for my server), it could cement the philosophical ties between them and elementary.

1¢ more

[qntnlq]

This aged poorly.

[reece146]

In spite of the Debian LLM vote I still think it is a worthwhile endeavour to break free of a corporately controlled base distribution with more and more dubious architectural changes to work around.

That said, perhaps Debian should be an option and not the assumed default. Finger memory can be short.

[ryonakano]

I looked up what at least we would need to do if we were to switch from Ubuntu to Debian.

Ecosystem

Launchpad (Size: XL)

We're using Launchpad to provides our packages for users; source package recipes, git mirror repositories, and PPAs. However, it only support building packages for Ubuntu series and not for Debian ones. So,

• we need to either:
  • seek if there is a free and trustworthy alternative service
  • create our own build and PPA server, which costs for power and maintainance

Platform layer

platform repository (Size: L)

The current default branch resolute which is for OS 9 is based on resolute branch of ubuntu-seeds repository. We need to:

• Research if there is an alternative repository for ubuntu-seeds repository
  • If exists
    • Rebase our platform repository on it
  • If not
    • Research how Debian handles preinstalled packages
    • Add our customization in accordance with the Debian's way

seeds repository (Size: L)

Similar to platform; The current default branch resolute which is for OS 9 is based on resolute branch of ubuntu repository. We need to:

• Research if there is an alternative repository for ubuntu repository
  • If exists
    • Rebase our seeds repository on it
  • If not
    • Research how Debian handles preinstalled packages
    • Add our customization in accordance with the Debian's way

os-patches repository (Size: M)

• get-latest-version.py
  • Modify to a way that does not depend on Launchpad
• Original source branches (e.g. bash-resolute)
  • Import original source codes of packages we'd like to patch from Debian not Ubuntu
• Patch branches (e.g. bash-resolute-patched)
  • Port our customizations to the original source branches imported from Debian

appstream-data repository (Size: M)

• build.sh
  • Modify to a way that does not depend on Launchpad

os repository (Size: S)

• etc/auto/config
  • Switch MIRROR_BINARY_URL and MIRROR_BINARY_SECURITY_URL to the Debian's ones
• etc/config/archives/*.list
  • Replace ubuntu with debian in the URL of our PPAs
• etc/terraform-*.conf
  • Switch MIRROR_URL to the Debian's ones

metapackages repository (Size: S)

• update.cfg
  • Replace the Ubuntu series with the Debian's ones
  • Replace the URL of our PPAs

Application layer

Desktop app repositories (Size: XL)

• .github/workflows
  • Investigate which workflows need to switch runs-on from Ubuntu containers to Debian ones
    • We're currently using Ubuntu runners in the most workflow files of GitHub Actions: see the search result
    • For example workflows that natively builds an app on a Ubuntu runner would need to be switched to Debian
  • Switch runs-on from Ubuntu containers to Debian ones in the workflows investigated above
• debian/changelog
  • Change release target series (e.g. granite9 (9.0.0) resolute; urgency=medium to granite9 (9.0.0) forky; urgency=medium)

settings-daemon repository (Size: M)

• src/Backends/UbuntuDrivers.vala
  • Modify not to depend on ubuntu-drivers command that is not available on Debian

Websites

website repository (Size: S)

• index.php: https://elementary.io/en
  • Remove link to Ubuntu Security Notices
• open-source.php: https://elementary.io/en/open-source
  • Replace Ubuntu entry with Debian
• docs/learning-the-basics.md: https://elementary.io/en/docs/learning-the-basics#updating
  • Remove mention about Ubuntu Security Notices

[qntnlq]

Well that's an awful lot...
Thank you @ryonakano for clarifying what would need to be done !

[reece146]

Nick at The Linux Experiment put out a video about Tuxedo's beta of the rebase on Debian.

https://www.youtube.com/watch?v=FKpZO1bZtYI