Article 79WSJ Ubuntu Core Desktop targets 28.04 debut, with default status in its sights

Ubuntu Core Desktop targets 28.04 debut, with default status in its sights

by
Liam Proven
from www.theregister.com - Articles on (#79WSJ)
Story Image

Canonical is targeting Ubuntu 28.04 LTS for the release of its immutable Core Desktop, with a stretch goal of making it the default download for Ubuntu 30.04.

Canonical's VP of engineering, Jon Seager, set out the timetable in his opening keynote at Ubuntu Summit 26.10. Yes, we know, we have heard this before: we reported on his plans last month. But this time, there's the beginning of a release schedule.

Seager said that Ubuntu Core is already out there in tens of thousands of IoT devices - which seems like a modest estimate to us, and to him too, as he quickly added "maybe more." Despite some noise a few years ago, though, what's still missing is the desktop version.

Seager's slides listed three bullet points: that Core Desktop will be "immutable," "composable," and "attestable."

As he put it, they are "taking the whole of the Linux operating system, and cramming the whole thing into snaps." Acknowledging that this is going to be wildly controversial, he said: "This is one for the Redditors."

He's right. There is a lot of resistance to snap, and in direct response, advocacy of the Red Hat-backed Flatpak instead - but we have seen almost none that is based on either system's technical merits. Much of that opposition concerns claims of one being more open than the other. We've looked at how both systems work, and our impression is that the implementation of snap is cleaner, simpler, easier to use, and more comprehensible. Snap does depend on systemd, which Flatpak does not - but only for now. Flatpak NG is in development and will also need everyone's favorite init system.

Ubuntu Core's immutability is not new technology: even Core 16 is now a decade old. It has supported in-place version upgrades since Core 22.

We have reported before that Canonical remains strongly committed to its packaging format, and that commitment seems stronger than ever. Seager continued: "Never has a model like snap confinement been more relevant... we're finally in a position where we can bring this to the desktop."

The core idea of an immutable OS is that "much more of the system is not writable at runtime," meaning that it "can't be messed with once the system has booted." He talked about the same three use cases we discussed last month. The "browser-centric machine" is the company's analog to Google ChromeOS, and he said that it's essentially ready now, suggesting that everyone listening probably has a family member whose computer is basically just a way to use a web browser.

The second was the dedicated games machine, which we interpret as a response to Valve's SteamOS. This edition, he said, would "boot straight to Steam," with "just enough of a shell to launch games." There's also an immutable "development workstation," and here, he offered the example of developers who need to work on some long-unsupported hardware or software, for whom an immutable OS means that you "can't brick your system by installing some weird old dependencies, unmaintained for 20 years."

Another important benefit is "atomic, transactional updates," which he said "brings enormous stability." The terminology comes from the database world: an atomic operation succeeds as a whole or is not applied. The aim is to avoid leaving the system stranded between incompatible package versions, with rollback providing a route back if an update causes trouble.

The "composable" part means assembling the OS from separately packaged components, from desktop environments to underlying services such as PipeWire and fingerprint-reader support.

Here, Seager used an example that spoke to us. He asked: "Ever installed something and watch as it spewed stuff all over your system?" Yes, we have, many, many times. You install one innocent-seeming little component, and it brings in a database, three new server components, and about half of a different desktop environment. We mentioned an example when we described how to customize Xfce a couple of months ago. That time around, in Debian, we found it was safe to install all packages starting with "xfce4-" except "xfce4-dev-tools." Most are tiny panel controls taking just hundreds of kilobytes, but the "dev-tools" package also pulls in several hundred megabytes of compilers, linkers, and other programming tools, totally unrelated to desktop layouts - and once installed, they're surprisingly tricky to remove again.

This is part of the appeal of a composable immutable OS: adding a component need not mean tangling its dependencies into the base system. But easier housekeeping isn't the only attraction. Canonical kept snap packages from its cancelled phone and tablet OS because snaps enable it to maintain a single package of, for instance, Firefox, which will work on all of its different OS versions that are still getting updates. Moving such important external dependencies into isolated snap packages makes it easier and safer to upgrade from one version of the OS to another. Complex sets of dependencies are the main point of failure of OS upgrades, especially when the OS has been used for a while.

This is very much not a solved problem. For instance, Debian version upgrades can be fraught - especially when the installation has decades of history behind it. Debian does at least support version-to-version upgrades. For many years and versions, RHEL did not - which is why we highlighted AlmaLinux's tooling for the task. Zorin OS added version-to-version upgrades in 16.3 just a few years ago. Linux Mint introduced its own upgrade tooling a few years earlier, although it still imposes restrictions.

It's a tough issue to crack. Major enterprise vendors discourage it, and phone OSes are all immutable. Helping with this problem is one of the main reasons snap packaging exists, while Flatpak is almost entirely irrelevant to it.

Seager also stated that Core Desktop would support "multiple desktops from the start." This will not be a GNOME-only project. Indeed, he specifically mentioned a KDE Plasma version, saying: "Plasma is great. People love it. We will have a KDE edition, with rolling Plasma. We will deliver upstream Plasma, as fast as we can." That sounds like a very welcome counterpoint to the LTS version of Plasma 6.6 in Kubuntu 26.04 whose announcement we covered back in August.

For the "attestable" line of the slide, he claimed that Core Desktop was "evolving the way we think about 'trust' on a machine. We have TPM, secure boot, TPM-backed Full Disk Encryption... but they're all bolt-ons." Core Desktop, he added, could "tie this into a hardware root of trust," delivering "verified system integrity." Canonical revived its TPM-backed full-disk encryption system back in 2025, and it arrived in Ubuntu 26.04. Back in January, systemd creator Lennart Poettering launched his startup Amutable, and earlier this month published a technical overview of its new immutable Linux. We suspect Canonical is watching closely, and that this is a hint in that direction.

We were also amused by Seager's mention of another rather niche distro: he said that Core Desktop would deliver "all of the joy that people derive from NixOS, for those with that perverse habit - but for everybody."

Daily preview installation images of Core Desktop will, apparently, be available very soon, in both GNOME and KDE flavors, and he warned that "there be dragons." (That may be a veiled nod to one of founder Mark Shuttleworth's other companies.)

After the promises made in 2023, we'll be watching that timetable closely. "I believe this is the future of Linux," Seager said, promising "a final version by 28.04, released alongside Ubuntu," adding that a "very stretchy stretch goal was by default for 30.04: download from ubuntu.com and you get Core Desktop - unless you ask for the other thing."

For those alarmed by that prospect, he had reassurance: "Ubuntu as you know it today is not going anywhere. There are no plans for the existing distro to go away."

Core Desktop wasn't even the main subject of his keynote. They followed a lengthy excursion into formal logic: definitions, proofs, invariants, and designing software whose behavior can be checked against precisely stated rules. The accompanying slides showed the logos of organizations working in this area, such as Hegel, Antithesis, and the Lean programming language. For now, an entire OS is much too big to formally verify, but subcomponents may be doable - and Canonical is starting with snapd.

He also expressed his satisfaction with the Stonking release, highlighting that it advanced Canonical's move toward Rust-based components, including Rustls and upki - noting that this proved very timely in the light of recent events. The release also delivered the newest available kernel, OpenSSL 4 - which he called a "huge change" - and official RISC-V images. He also called out the Myna speech-recognition feature, saying that it was all in-house, designed and shipped in six months by a team of "about two and a half people." It will form part of the company's plans for what it calls a context-aware desktop. (R)

External Content
Source RSS or Atom Feed
Feed Location http://www.theregister.co.uk/headlines.atom
Feed Title www.theregister.com - Articles
Feed Link https://www.theregister.com/
Reply 0 comments