First and foremost, a heartfelt thank you for the spirited discussions about the Ubuntu desktop ISO and our vision for it. We value your input and are grateful for your ongoing engagement.
Simplifying Installation: Where We Are Now
As many of you are aware, we had been exploring the consolidation of the ‘Full’ and ‘Minimal’ installation options into a single ‘Default’ choice. However, due to a key team member recently moving on to new opportunities, we can’t fully implement our vision at this time. Accordingly, we’re adjusting our plans. For the 23.10 release, we plan to introduce this new ‘Default’ option while retaining the existing ‘Full’ option. Though limited, this change should offer some valuable insights that will help further inform our ongoing dialogue.
The Bigger Picture: Ubuntu Provisioning Overhaul
At the same time, we’ve been diligently working on a comprehensive overhaul of Ubuntu provisioning, targeting everything from ISO downloads to day-to-day usage. Given the current circumstances, however, we’re adjusting our objectives for the 23.10 release and will re-evaluate our roadmap for 24.04 in due course. When this work is presentable, I will of course share it.
Thank you, I really appreciate your attitude and approach to the matter.
I’d love to see a default iso that satisfies everyone’s needs. But I don’t think that this “universal” iso exists.
So I think ubuntu should gradually adapt to the needs of each user starting from the full installation (which with the large number of software it is equipped with is able to respond in the first instance to the needs of almost all new users). In practice Ubuntu could suggest the removal of software that hasn’t been used for 2/3/X months in order to lighten the system and save data (and battery) for updates. (This is a substantially opposite approach to what has been proposed so far: the system suggests to the user what to install. I would propose instead what to delete after verifying what is not being used. Probably this could lead to the telemetry minefield.)
Until now I don’t see a FULL option. … waiting. Thanks
I have no problem to install all the missing apps I need, but inexperienced would find the system unsuitable for day-to-day work
I have no problem to install all the missing apps I need, but inexperienced would find the system unsuitable for day-to-day work
As someone who works with those inexperienced users regularly I share the same concern but the hope is it will be mitigated by having the two options well explained in detail before the user chooses one.
Also I think the numbers will tell us soon enough how valuable each option is. Not the percentage of users choosing each option but whether we can make the Default (Minimal) option sufficiently smaller than Full so as to justify having two options.
At the moment Ubuntu offers many packages twice. Once as a DEB, once as a Snap. This causes one of the two variants to be poorly maintained.
Now, someone installing LibreOffice and Thunderbird with the default option will get a different flavour of the package (snap) than someone installing the full option (DEB). Canonical should definitely unify that everyone gets the same version of the app, either snap or DEB. Delivering different variants depending on the installation variant is unprofessional.
I think the main reason why Canonical wanted to get rid of the full option is because of the high maintenance and support requirements of this option. A good compromise would be that Ubuntu continues to offer the full option as an option, but this should be nothing more than a wrapper that preinstalls various snaps (ubuntu-desktop-full.deb = snap install libreoffice thunderbird games etc.). This could save Canonical a lot of maintenance work. I can also understand Canonical here, since even a giant company with tenfold sales like Red Hat can no longer afford to provide LibreOffice as a classic package.
But then Canonical finally has to commit to maintaining the snaps properly. (The existing snap maintainers are doing their best, hence kudos to them, but Canonical should hire some full-time employees just to handle the maintenance of important snaps). I have not yet managed to start a running VM with the Gnome Boxes Snap, but the DEB works perfectly. The GIMP Snap hasn’t been updated in aeons, the GIMP 3 Beta Snap won’t start at all, as will the Gnome Drawing Snap. The JOSM Snap is unusable compared to the Flatpak, does not support HiDPI and no GTK theme probably due to an old Java version. The LibreOffice Snap starts much slower than the DEB version.
Firefox Snap has proven that Snaps are technically on a par with Flatpaks. The prerequisite for this is that the snaps are properly maintained. So that Ubuntu can pre-install various snaps, they simply have to be properly maintained.
I’m not sure if i fully understand your idea, but other distros already do this. If an user wants to remove only one of these packages (e.g thunderbird) the package manager usually requests to remove the entire “ubuntu-desktop-full.deb” packages bacause that package is a dependency. This causes the system to break. This may be an interesting option, but it’s dangerous if not well implemented.
I don’t remember if Ubuntu’s installer already has this option, but in my opinion the installed should have a page where the users can select the additional software they want. Not all the software, like a Synaptic page, only some categories. Eg. “Office suite” (e.g libreoffice), “Internet” for additional internet packages (e.g thunderbird). 4-5 categories that the user can thick if interested.
This doesn’t work if you want to shrink down the ISO size
Τo be honest, while a bit irrelevant with the whole discussion, I believe they should move all the way to snaps. What’s the reason to package Thunderbird and LibreOffice as snaps but offer debs in the install. Actually no, I can see reasoning behind it, but I would embrace them dropping these debs for good quality snaps.
I suppose the question I have is: knowing that an install that doesn’t come with more than a couple tools and a web browser is inadequate and hostile to users without some kind of app store experience that doesn’t yet exist and there are no solid plans for it to exist in 24.04 LTS (yet! I know you’re working on it)…
Why make the default Ubuntu desktop experience inadequate and hostile by default? Why not simply adapt the minimal install and change the default after we know it’s ready, instead of now that we know that it isn’t ready at all?
This new plan still makes it a lot more difficult for me to recommend Ubuntu to new users for the next year, especially since I tend to recommend the latest interrim release if someone needs Ubuntu a couple of months before the next LTS, to maximize their support time and minimize the upgrade’s impact on the user experience.
I like the direction where this is going, and find the changes introduced so far in ubuntu-desktop and ubuntu-desktop-minimal to my taste.
A few suggestions related to the current dependencies of this packages (in no particular order). Because I see some a lot of packages there that I understand were accumulated over time:
gnome-terminal: Can it be replaced by gnome-console? I understand it’s intended to be the new default for Gnome.
printer-driver-*: Currently, it installs the 9 printer-driver packages. Although I have no use for these, I understand that you want to include them in a standard installation. But can these 9 packages be replaced by printer-driver-all or printer-driver-all-enforced. This way we wouldn’t have to remove the 9 packages by hand, but just one.
fonts. You install fonts-dejavu-mono, fonts-liberation, fonts-noto-cjk, fonts-noto-color-emoji, fonts-noto-core and fonts-ubuntu. Personally, I don’t love to have a very base large of fonts by default, and some of these are already metric equivalent. Are all these required by the default installation? I would suggest to just install the fonts required by the default interface and software of Ubuntu.
cups. In addition to cups, you install explicitly install cups-bsd, cups-client and client-filters. At least, the latest two are already dependencies of cups and can be removed of this list.
update-manager: Already has a update-notifier as a dependency. The second is not needed in this list.
ibus: Already has ibus-gtk and ibus-gtk3 as dependencies. These are not needed in this list.
language-selector-gnome: Already has language-selector-common as a dependency. The second is not needed in this list.
xorg: Is stated as a dependency. Is it still required?
alsa: alsa-base and alsa-utils. Same as above.
I have not reviewed all lib* packages, but I guess at least some of them do not require to be a direct dependency of ubuntu-desktop.
As you can see, in general, I think the list of depends and recommends of these packages can be trimmed down to a tidy and ordered list.
We copied the iconic purple background from Terminal to Console for 23.10. We are unable to switch the default terminal app to Console until our screen reader works with Console.
As far as I am aware, no other mainstream GNOME distro has switched to Console yet. Ubuntu therefore might be one of the first. There has been some hesitation about the lack of customization options compared to Terminal. Console 45 now has a very minimal preferences dialog, but it would be helpful to identify which are the most urgent customization options needed in a basic terminal app.
It’s actually hard to tell which fonts are required until they go missing. I know for sure that fonts-ubuntu (or fonts-ubuntu-classic) is critical to the desktop, and fonts-noto-color-emoji is where all our emojis live.
But also automatic font fallback occurs when your primary font is missing a Unicode character it’s been asked to display. This occurs for example with the colon in the gnome-shell panel clock because it’s designed to use an uncommon character that Ubuntu (most?) fonts don’t implement.
xorg: Is stated as a dependency. Is it still required?
Yes Xorg is critical to supporting Nvidia systems and many types of virtual machines. I hope we can work to make it optional in future releases not least because Ubuntu Core Desktop isn’t expected to provide Xorg sessions (but does support X apps via Xwayland).
I kind of read the request that dependencies that are pulled in anyway by a seeded package should not be explicitly listed in the seeds, not that @mlorenzo wants us to actually remove the packages from the install…
(But perhaps I’m misinterpreting)
Thank you, @ogra. Yes, that was exactly my point. I was just talking about first-level dependencies.
Compare, for example the list of packages that you have as a first-level dependency of ubuntu-desktop-minimal with the ones that Debian has in gnome-core (which can be a good equivalent): Debian – Details of package gnome-core in bookworm
While it’s true that there are still some libraries there, you mostly find applications. That should be the point of the first level dependency hierarchy. Every library or essential component that you need to make explicit at this level is a symptom of incongruences in lower levels of package’s dependencies.
Using the example stated above, doesn’t gnome-shell pull the dependences required to have audio? Do you need to declare explicitly alsa as a dependency of the desktop? The response is no, you don’t. Although, in this particular case it would probably be wise to replace alsa-base, alsa-utils, pipewire-pulse, wireplumber and libspa-0.2-bluetooth for just pipewire-audio
Thank you all for your replys and you interest on improving the desktop.
I agree our first level dependencies are sometimes excessive, and noticed such an issue in gnome-shell recently. I suggest the best way forward with these is to log a bug for each one.
Today in 2 different partitions I installed the Minimal system and the Default/Full from the ISO Daily amd64 (20230915).
Both are excellent, the Minimal is perfect for an expert user who will install what he needs, the full for an inexperienced user who finds everything he can need. Thank you
