Why I Left Ubuntu Desktop — and Picked Debian Over Ubuntu Server

· 9 min read

I ran Ubuntu on my workstation from Hardy Heron in 2008 to about Jammy in 2022. Fifteen years. It was the first Linux I trusted on servers I cared about, the first one that made hardware support feel solved, and for a long stretch it was the obvious answer to “what Linux should a sane person install?”

I don’t run it on my desktop anymore. I don’t run Ubuntu Server either — though that’s a personal-taste call and not a knock on the product, which is still a perfectly good distribution that I’d happily recommend to most people. Ubuntu has split in half: the server side is still strong, the desktop side has gone somewhere I’m not willing to follow, and the two halves deserve very different treatment. Especially because we just got Ubuntu 26.04 LTS “Resolute Raccoon” on April 23, and the headlines are mostly cheerful about features that don’t fix the things that actually drove me away.

The good news first — 26.04 LTS is a real release

Credit where it’s due. Ubuntu 26.04 LTS has serious things in it:

  • Linux kernel 7.0, which means real Intel Panther Lake NPU support, the latest io_uring work, and the kernel-side fixes that actually move performance on 2026 hardware.
  • GNOME 50, X11 removed from GDM entirely — desktop sessions are Wayland-only now. (For my taste this matters more for the principle than the product, because I’m on KDE. But Wayland-only as the default is the right call in 2026.)
  • sudo-rs is the new default sudo. The original sudo is renamed to sudo.ws. This is the largest Rust-rewrite-of-a-core-utility to ship in a mainstream distro and a real bet on the next decade of CVE-resistance.
  • NVIDIA CUDA in the official repos for the first time. No more third-party PPA dance for AI workloads.
  • TPM-based full disk encryption by default on fresh installs.
  • OpenJDK 25, Rust 1.93, LLVM 21, .NET 10 — all first-class.
  • cgroup v1 is gone from the kernel build. v2-only. The right decision, four years late.

If you’re running Ubuntu Server, this is a strong LTS. Five years of support to 2031, ten with Ubuntu Pro. The technical foundation is genuinely good.

So why am I writing the rest of this post?

The four things that broke my desktop trust

1. Snap, as a product and as a posture

Snap is Canonical’s universal package format. The idea — sandbox an application, ship it as a self-contained bundle, auto-update it — isn’t wrong. Flatpak does the same thing and I run Flatpaks happily. The problem is everything around Snap on Ubuntu, in three specific ways:

  • The Snap Store is proprietary. Only Canonical can operate one. This is, technically and philosophically, an open-source product with a closed distribution monopoly bolted on. Flathub is run by a non-profit; anyone can stand up a Flatpak repository. Snaps don’t work that way.
  • You cannot disable auto-updates, only defer them. A package manager that won’t let me say “no, not now” is a package manager that’s making a different deal with me than I want to make.
  • Sandbox boundaries break common workflows. The Firefox snap — more on this in a moment — has been the canonical (pun acknowledged) example for four years now: broken KeePassXC integration, broken YubiKey, broken printing for some configurations, dramatically slower cold start, and a profile path of ~/snap/firefox/common/ that quietly diverges from where firefox used to put things. None of these are unsolvable, all of them are friction every single time.

2. The silent .deb-to-snap swap

This is the one that, more than anything else, changed how I felt about Ubuntu.

On 22.04, apt install firefox started installing a tiny .deb shim that itself installed a snap. Same command, totally different delivery mechanism, no flag to indicate the change. By the time you noticed your browser had moved to ~/snap and was sandboxed in ways that broke your password manager, the migration had happened.

The same pattern has since spread: Chromium snap-only, Thunderbird snap-default since 24.04, and a growing list of desktop tools where the .deb is a placeholder for a snap. Mozilla ended up shipping their own .deb repository because the snap experience was bad enough to threaten their relationship with Ubuntu users.

This is not a technical complaint. It’s a trust complaint. When the same command produces a fundamentally different artifact than it did last month, without an opt-out and without a heads-up, the distribution is no longer telling me what it’s doing. I will accept opinionated defaults from a distro. I will not accept opinionated defaults that look like the old defaults.

3. Ubuntu Pro in my terminal

apt update now prints a paragraph about Ubuntu Pro most of the time. The MOTD on a fresh Ubuntu Server shell prints another. The Software Updater nags. The notifications nag.

I understand the security argument — “ESM patches exist, they cost money, we should tell you.” I’d accept it once, in apt update, behind a --silence-pro flag, the way every other vendor handles this. I will not accept it every time, in every shell, on every machine, for the rest of the LTS lifecycle. It is — literally, structurally — advertising in my terminal. That’s a line.

4. The cumulative trust drift

None of the above is a single fatal flaw. Ubuntu is still a working desktop. You can remove snapd, install the Mozilla .deb, mute the Pro nag with pro config set apt_news=false, and run a perfectly fine 26.04 desktop.

The point is that I shouldn’t have to. Every Ubuntu release for the last four years has added one more thing I need to undo on first boot. Every release has shifted the default slightly further from “this distro is for me” toward “this distro is for the corporate fleet customer Canonical hopes to sell Ubuntu Pro to.” The drift isn’t dramatic on any one release. It’s been steady, and steadily in the same direction, since 22.04.

Server-side, Ubuntu is fine — but I run Debian

Here’s the honest accounting on the server question.

Ubuntu Server in 2026 is, by most objective measures, an excellent server distribution:

  • It’s the cloud default everywhere — AWS, GCP, Azure, Hetzner, DigitalOcean, Linode, OVH. Either the default image or a co-default with Debian. Cloud-init scripts assume it. Half the Ansible inventory you’ll inherit at a new gig assumes it. The ecosystem gravity is real and it cuts in Ubuntu’s favour.
  • Server-side, the snap problem barely exists. You’re not running Firefox on a headless box, and snapd only really shows up if you go looking for it (or use LXD, which is genuinely good).
  • Five years of LTS support, ten with Ubuntu Pro. At fleet scale that pricing is defensible.
  • Kernel 7.0 + sudo-rs + cgroup v2 is a legitimately strong base.
  • unattended-upgrades works out of the box, cloud-init integration is polished, the documentation is thorough.

If you’re picking a server distribution in 2026 and you don’t already have a preference, Ubuntu Server is a perfectly reasonable answer. I’d rather you run it than half the alternatives.

So why do I run Debian on my own servers? Three reasons, in declining order of how much they matter to me:

  • Vanilla, no commercial overlay. Debian doesn’t print Ubuntu Pro ads in my MOTD. It doesn’t push me toward a paid support tier. It doesn’t include snapd by default. When I run apt update on a Debian box, I get package updates and nothing else. That’s the agreement I want with a package manager.
  • Stability that’s genuinely glacial in the good way. Debian 13 “Trixie” shipped in August 2025 with a freeze process you could set your watch by. The packages are conservative, the security team is experienced, the upgrade path is well-trodden. For boxes that need to keep running for three to five years between major upgrades, that’s exactly the trade I want.
  • Community governance. No CEO-driven product strategy, no quarterly revenue targets to hit. If I bet on Debian for a side project I expect to still be running in 2031, the distribution is going to look broadly similar to what I started with. With Ubuntu I can’t say that with the same confidence — the trajectory has changed twice in the twenty years I’ve been paying attention.

This isn’t a moral position. It’s a working preference. The reader who picks Ubuntu Server for cloud image availability, LTS contracts, and operational predictability is making a fine choice. I’m just not the one making it on my own infrastructure.

My NetBird coordination server? Debian 13. The Hetzner CX22 that holds a couple of side projects? Debian 13. The little ARM box running Home Assistant? Raspberry Pi OS — which is Debian with a hat on. There’s a pattern, and the pattern is “boring, vanilla, no surprises.”

What I run instead on the desktop

Fedora 44 KDE. Has been for a while now. The trade-offs of running a six-month-release desktop are real — more frequent upgrades, less corporate predictability — but the trade is:

  • No silent package-format swap.
  • No proprietary store as the default backend.
  • KDE Plasma 6 as a first-class desktop, not an afterthought.
  • Flatpak as the sandbox story, Flathub as a community-governed store.
  • A clear path to atomic (Kinoite, Silverblue) if I ever want the rebase-rollback model, with Toolbx and Distrobox handling the userland.

Fedora has its own strong opinions — Wayland, PipeWire, systemd integration, the whole modern Linux desktop stack — and I happen to agree with all of them. More importantly, those opinions are technical, not commercial. The distro isn’t trying to sell me something every time I open a terminal.

Who should still run Ubuntu Desktop

I want to be fair here. Ubuntu Desktop is the right choice for:

  • First-time Linux users. The hardware support, the install flow, the giant pile of online documentation, the ten years of Stack Overflow answers — all of that still adds up to the easiest on-ramp.
  • Corporate / education fleets. Snap’s auto-update is exactly what an IT department wants. Ubuntu Pro is exactly the support contract a procurement team understands.
  • People who don’t care. And I mean that without sarcasm — a lot of people just want a browser, an email client, and a reasonable file manager. Ubuntu Desktop delivers that, snap-or-not.

For me, none of those describe my life. I am the user who notices the silent .deb-to-snap swap, who reads /etc/motd because something looks off, who profiles cold-start times because a slower Firefox bothers me on principle. For that user, the Ubuntu desktop of 2026 is not what it was in 2014.

The summary, in two lines

Ubuntu Server in 2026: still a strong product, especially at 26.04 LTS. I’d recommend it to anyone who hasn’t already settled on something they prefer. For me, that something is Debian.

Ubuntu Desktop in 2026: the relationship has moved on without me. No hard feelings. Just no more apt install firefox and hoping for the best.