On the use of the Calamares installer in Ubuntu

Tue May 6 00:47:34 UTC 2025

Hi Everyone:

I am the Technical Product Manager of Kubuntu Focus systems. I agree with
Aaron that we would see little benefit from switching to the Flutter
installer in Kubuntu 26.04 LTS, yet we would expose the Kubuntu project to
significant costs and risks.

I am writing this from the perspective of an OEM that contributes in many
ways to the Kubuntu project. For some background, we have a primary
objective of shipping this official Ubuntu flavor integrated, validated,
and supported on hardware, similar to what you would expect with a MS
Surface or Google Pixel device. For example, we revalidate all models sold
in the last 3 years on every kernel update, and on every major GPU and
desktop update across hundreds of KPCs.  See
https://kfocus.org/land/business for more details.

Because we ship Kubuntu LTS, we are a substantial supporter of the Kubuntu
project. We arrange product reviews that feature Kubuntu to the likes of
Ars Technica, ZDNet, TechRadar, HowToGeek, and LinuxInsider. We also
developed quite a bit of code, configurations, and artwork for Kubuntu,
especially the LTS releases. We donate laptops to Kubuntu developers; pay
for a corporate-level KDE patronage and sponsor KDE bug fixes for Kubuntu;
extensively test and report on desktop, kernel, and drivers updates over
hundreds of KPCs (as above); and create and upstream fixes to Kubuntu, the
kernel, and Ubuntu proper. These activities certainly are symbiotic with
the Kubuntu team, and improvements often benefit all Ubuntu flavors.

As with 24.04 LTS, we plan to be a significant contributor to 26.04 LTS.
What follows are the primary concerns I have with a full or partial move to
the Flutter installer:

=====

1. Integration will likely be expensive and risky: The Kubuntu team would
need to replace a robust and fully developed installer with one that is not
yet feature-complete, tested, or proven for our OEM use cases. This exposes
Kubuntu to substantial technical and schedule risks.

2. It would divert substantial resources: Any time and expense we would
incur on developing a new installer would proportionally diminish the
resources available to the core Kubuntu developers.

3. It would be far more expensive than you think: Replacing the Calamares
installer would require a full audit, analysis, and retooling of the entire
full installer support stack. This is many multiples of cost beyond just
reimplementing the installer alone.  Required changes would include at
least installer hooks and scripts, the hardware optimization subsystem, OEM
assembly scripts and procedures, validation scripts and procedures, CSR
procedures and training, various references throughout our online docs,
system tools, and the annotated images and text for the end-user
documentation found at https://kfocus.org/wf/reinstall.

When Aaron says this takes hundreds of hours, he’s not exaggerating; we
have the time logs from the last installer change to prove it. And that was
with a feature-complete and highly tested installer. ALL resources spent on
this initiative would be diverted away from core Kubuntu development.

4. It would put other critical work at risk: the 26.06 LTS upgrade from
24.04 LTS changes the display server from X11 to Wayland, it moves from Qt5
to Qt6 and a move from KDE Plasma 5 to KDE Plasma 6. These require
substantial rework across the entire product support stack: install hooks,
hardware subsystem, assembly, validation, CSR, system tools, and online
docs. Adding a new installer would imperil this effort.

=====

These issues are not specific to us; they affect any OEM that needs to
integrate to the installer to unlock hardware features. Other OS vendors
often support OEM-sensitive systems like this for decades because they
realize that even a simple change can have a massive ripple effect that
invalidates documentation and breaks many support systems.

I hope this response illustrates the full cost of changes to an
OEM-critical component, and how changes like this can unwittingly encourage
OEMs to move to their own, non-official Ubuntu derivative. I also believe
it illustrates an attitude and a way forward that can encourage OEMs to
participate in the improvement of the Ubuntu flavor they ship, as we have
done.

Sincerely, Mike

Michael Mikowski
Technical Product Manager, Kubuntu Focus
415-418-4731 09:00-18:00 EST | kfocus.org

On Mon, May 5, 2025 at 11:47 AM Aaron Rainbolt <arraybolt3 at ubuntu.com>
wrote:

> This is related to an email from Jon [1] suggesting that the all Ubuntu
> flavors use the Ubuntu Desktop Installer. I discussed this some with
> Jon in an off-list email chain, this is me bringing some of the
> arguments I made from that thread and elsewhere into public so they can
> be discussed more thoroughly. Based on my discussions with multiple
> developers in the Lubuntu and Kubuntu development circles, I believe
> both Lubuntu and Kubuntu are opposed to this change, and will try to
> explain why here.
>
> Lubuntu has, for the last thirteen cycles, been shipping the Calamares
> installer as our default installation system. This is a very mature,
> very well-established, lightweight, and functional installer that has
> worked without severe issues for Lubuntu for many years now, save for
> the last cycle where some unfortunate development practices resulted in
> serious bugs very close to release (the details of which I won't go
> into for privacy reasons). Kubuntu and Ubuntu Unity have also been
> using this installer for the last three cycles, and at the time of
> switch we were unaware of any pressure to select the Ubuntu Desktop
> Installer instead of Calamares. Kubuntu analyzed both installers when
> switching from Ubiquity and made an intentional design decision to
> switch installers to Calamares. Canonical's request here is not a usual
> system-wide change such as when Firefox was switched to being a Snap,
> or the current plans to swap out GNU coreutils for uutils. The request
> is to override an intentional, well-established and long-standing
> design decision made by multiple flavors without Canonical's previous
> expression of displeasure with this decision.
>
> There are several problems with this request that make me opposed to it:
>
> * The request fundamentally misunderstands what flavors need to
>   maintain their identity and provide a desirable user experience.
> * The speed with which Canonical is suggesting making the change is
>   inconsiderate of the needs of flavors and those who depend on them, to
>   the point of neglectful.
> * The effort needed to fulfill this request will be a severe drain on
>   the development resources of multiple flavors to replace something
>   that isn't broken, against our will.
> * While less problematic than the other points, I do not believe the
>   "request" is being made in good faith, despite having intentionally
>   assumed good intentions when initially going into the discussion.
>
> I'll elaborate on these points below, copy-pasting each point before
> backing it up.
>
> To begin with, the request fundamentally misunderstands what flavors
> need to maintain their identity and provide a desirable user experience.
>
> One of the points brought up in the discussion between me and Jon in
> our off-list discussion is:
>
> > Regarding package support and archive presence: tools like Flatpak and
> > Calamares are in **universe**, which, by design, includes a broad
> > range of software that may not be core to Ubuntu’s identity. To be
> > clear, I’m not suggesting removing Calamares from the archive—only
> > that it shouldn’t be the default installer for an Ubuntu Flavour.
>
> This statement is problematic for the Ubuntu flavors, because it
> confuses the identity of Ubuntu as a whole with the identity of the
> individual Ubuntu flavors. Packages in `universe` may not be core to the
> identity of Ubuntu Desktop and Ubuntu Server, but that is because
> Ubuntu's policy is to only allow applications into Ubuntu Desktop and
> Ubuntu Server if they are present in the `main` or `restricted`
> repositories. Ubuntu flavors are not bound by this requirement (they
> can't be, otherwise Canonical would have to maintain all of the
> packages every single flavor uses), and as such they include a very
> large amount of software from the `universe` archive, much of which is
> extremely core to the identity of each individual flavor.
>
> The best example of this is desktop environments - every flavor with the
> exception of Edubuntu uses a desktop environment that is composed
> primarily of packages from the `universe` archive. The vast majority of
> these flavors (all of them except for Ubuntu Studio) are
> desktop-environment-centric - Kubuntu is Ubuntu with the KDE desktop,
> Lubuntu is Ubuntu with the LXQt desktop, etc. These desktop
> environments aren't just important, they form the heart of each
> flavor's identity inasmuch as those flavors are separate from Ubuntu
> Desktop. If KDE and Plasma were to be removed from the archive, Kubuntu
> would no longer exist. Even if the Kubuntu team were to pick some other
> desktop environment, whatever they would end up with thereafter would
> no longer be Kubuntu.
>
> Most of the flavors are not particularly attached to their installer.
> Lubuntu is different in this regard however - the installer has become
> a part of our identity as a flavor. The Lubuntu team has put in an
> immense amount of effort to make it work well for us, including
> theming it, configuring it, writing extensions for it, contributing
> new code upstream to Calamares where appropriate, and even
> developing an entire set of tools that implement an OEM installation
> mode similar to the OEM installation mode originally offered by
> Ubiquity. The Calamares installer forms a core part of Lubuntu's
> identity second only to LXQt itself. It has for years.
>
> Jon argues that Calamares shouldn't be the default installer for an
> Ubuntu Flavor. From the perspective of Lubuntu, this is as far "out of
> left field" as saying that LXQt shouldn't be the default desktop for an
> Ubuntu flavor. We would obviously be opposed to such a suggestion
> because it would be threatening the core of what Lubuntu as a flavor
> is. Suggesting the removal of Calamares is an equally alarming request
> and something we are all but entirely unwilling to do, to my awareness.
>
> In addition to the above, Canonical has been starting to show a trend
> of requesting flavors to stop shipping particular packages in
> `universe` that they have some opposition to for some reason or
> another. This is not something I believe the flavors should have to or
> sustainably can keep doing every time Canonical asks for it. We pick
> packages from the `universe` archive to become part of our flavors as
> needed to deliver the features we believe our users will want. Those
> packages, while not core to Ubuntu Desktop/Server's identity, are still
> a part of the Ubuntu operating system, and to our awareness we as the
> flavors are free to pick and choose from them to provide the features
> we need. Every time Canonical requests us to drop packages, it means
> lost work and overridden design decisions. If Canonical doesn't want a
> package to be part of the Ubuntu flavors, it should not be part of the
> Ubuntu operating system. If it is part of the Ubuntu operating system,
> Canonical shouldn't be telling flavors to drop those packages unless
> Canonical also intends on having the packages dropped from the Ubuntu
> archive (after proper discussion with the involved flavors).
>
> My second complaint with this request is that the speed with which
> Canonical is suggesting making the change is inconsiderate of the needs
> of flavors and those who depend on them, to the point of neglectful.
>
> The Ubuntu flavors have communities, projects, and businesses
> (including system OEMs) depending on the feature set those flavors
> provide. Incremental changes are to be expected, but massive changes
> such as an installer or desktop environment transition requires advance
> planning, consultation with the community, and a long transition period
> as the community requires (within reason). This is how Lubuntu switched
> from LXDE to LXQt years ago. It is how Ubuntu Desktop switched from
> Ubiquity to the Ubuntu Desktop Installer. It's not how some of the
> other flavors have switched installers (and indeed, Kubuntu's switch
> from Ubiquity to Calamares was rather abrupt), but at the time of those
> transitions the respective flavors did not have community members
> sufficiently dependent upon them to warrant a slower transition.
>
> Kubuntu currently has a vital community member, Kubuntu Focus, who has
> been signally strongly that they would be affected very negatively by
> an abrupt installer change, even if they were allowed to keep the
> Calamares installer for 26.04 LTS as suggested by Jon. Kubuntu's
> transition to Calamares was intended to be permanent until such a time
> as Calamares was no longer suitable for use as an installer for Linux
> distributions (i.e., if it were abandoned upstream). Kubuntu Focus
> invested well over 100 hours of work into transitioning their
> documentation, processes, and software to the Calamares installer
> as a result, after being advised that it would be *the* new Kubuntu
> installer.
>
> To be considerate of the flavors, Canonical should not have attempted
> to force the flavors to switch installers in a single development
> cycle with the option of maintaining two separate images for two cycles
> if needed. This change should have been brought up *well* before the
> intended transition date, with the understanding that flavors would
> likely have a substantial amount of pushback which would need to be
> taken into account, potentially resulting in the installer transition
> being NACK'd. To bring it up not even a whole six months before the
> desired transition date and only provide a year for dependent OEMs to
> switch as needed is inconsiderate and neglectful for Canonical to do,
> especially considering that there was no possible way for any flavor to
> know Canonical would request them to switch away from Calamares.
>
> My third complaint with this request is that the effort needed to
> fulfill it will be a severe drain on the development resources of
> multiple flavors to replace something that isn't broken, against our
> will.
>
> Calamares, as I explained earlier, is very mature and has been working
> well for Lubuntu for years (and for Kubuntu for a year). While the
> 25.04 release cycle did have some substantial issues with Calamares,
> this was because of a Lubuntu-specific plugin that was developed by a
> trusted developer in an inconsiderate and neglectful manner. Lubuntu
> has done a postmortem of this incident and believes we will be able to
> prevent it from happening again through a combination of more thorough
> testing well in advance of the release. Given that the Calamares
> issues in 25.04 were an anomaly and not reflective of Calamares'
> stability and usability, there is no reason for Kubuntu or Lubuntu to
> switch.
>
> To go further, switching to the Ubuntu Desktop Installer would actually
> degrade the experience of Lubuntu and Kubuntu, because the Ubuntu
> Desktop Installer's design and user interface simply does not fit with
> the user experience offered by Lubuntu and Kubuntu. It uses a
> combination of theming, style, and UI toolkit that makes it look
> starkly different than literally every other app on both Kubuntu and
> Lubuntu, with the exception of the firmware updater snap. The installer
> will also be the first thing or almost the first thing the user sees
> when booting a live ISO of the distro, so this will immediately give
> users a sense of inconsistency and "half-baked-ness", negatively
> affecting the flavors' perception and reputation.
>
> Calmares works, and it works well. The Ubuntu Desktop Installer
> provides *nothing* that Kubuntu or Lubuntu need, want, or are even
> interested in. It's about as useful to us as GNOME Text Editor (i.e.,
> we have an application that serves the purpose already and the
> alternative would be quite inferior in the context of our desktops).
>
> Lubuntu and Kubuntu have plenty of things we are trying to develop as
> it is, such as Wayland support (Lubuntu has yet to get a Wayland
> session working, though we very much want to have one and have been
> working towards that goal with others). An installer transition would
> take time away from those more critical development tasks and allocate
> it to a task we find essentially useless.
>
> My final complaint with this request is that I do not believe the
> "request" is being made in good faith, despite having intentionally
> assumed good intentions when initially going into the discussion.
>
> In the original email to the Technical Board, Jon wrote:
>
> > I intend to ask all of the flavours to transition to the new Ubuntu
> > Desktop installer by 26.04 LTS. Now that it's stable for Ubuntu
> > Desktop, and we have evidence from the flavours who've already
> > adopted it, I'm confident that this goal is achievable within the
> > timeframe.
>
> Later in the off-list discussion however, Jon wrote:
>
> > You mentioned that you'd like to see a document defining what
> > constitutes an Ubuntu Flavour. This is being actively worked on and
> > while it's not yet mature enough to share, the document will state
> > that to be an Ubuntu Flavour, the Flavour must use the Ubuntu
> > Installer. The Community Team is leading on this definition, and will
> > circulate something soon.
>
> And also:
>
> > To summarise:
> >
> > - We’re finalising a document to define what constitutes an Ubuntu
> > Flavour, and will share it soon.
> > - Lubuntu should adopt the new installer starting with the next
> > release.
> > - Kubuntu should adopt the new installer _by default_ starting with
> > the next release, but may also ship an alternate edition using
> > Calamares for **25.10** and **26.04**.
>
> I am aware that the first quote has since been amended to say that "the
> document would likely state that flavours must use the Ubuntu
> Installer." However, my point still stands - Jon initially portrayed
> this as an ask, which I take to mean that the flavors could reasonably
> say "no" if fulfilling the request would cause severe damage or was
> otherwise undesirable (as I am arguing for here). The follow-up emails
> indicate strongly however that this is not a request, but rather a
> command. The time frame in which Lubuntu and Kubuntu are supposed to
> transition to the installer was furthermore shortened by an entire
> release cycle, making an already difficult request much harder to
> fulfill.
>
> I initially entered this conversation assuming good intentions, since I
> believe that is what is best for the Ubuntu project. From the resulting
> conversation however, I believe my assumption of good faith was
> incorrect and that the flavors affected by this discussion are being
> coerced into switching installers despite indicating a strong desire to
> not switch, with rationale to back up that desire. I do not believe I
> need to elaborate on why I consider this to be a problem, and I very
> much hope I am simply misunderstanding things here. I'm happy to
> privately forward the email chain to members of the Technical Board if
> desired (I do not want to send them publicly because of a confidential
> detail related to the Community Council discussed briefly in those
> emails).
>
> I'm happy to discuss any of these issues further as needed. Thanks for
> taking the time to read this, and I hope your day is going well!
>
> [1]
> https://lists.ubuntu.com/archives/technical-board/2025-May/003016.html
>
> --
> Aaron Rainbolt
> Lubuntu Developer
> https://github.com/ArrayBolt3
> https://launchpad.net/~arraybolt3
> @arraybolt3:matrix.org on Matrix, arraybolt3 on irc.libera.chat
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/technical-board/attachments/20250505/572d44f0/attachment-0001.html>