Ubuntu 23.04 'Lunar Lobster' beta is here in all its glitchy glory Ubuntu "Lunar Lobster" is the middle release between the 2022 and 2024 long-term support releases so this is when Canonical tries out some new stuff. According to Canonical's release schedule, Lunar should appear on April 20. Perhaps that will give the developers time to flesh out some of the sections that just read "tbd" in …
I suspect most people only ever install non-LTS Ubuntu by mistake, and it only take a few months to find out why it was a mistake.
In the old days you might get some importantfix but now base Debian is faster Ubuntu nonLTS releases add little (for desktop Users)
Since Snap nonsense, Debian has become much simpler than Ubuntu
Debian Bookworm will have LXD natively packaged, whereas Ubuntu only provides it as a snap.
As this is the last thing I use that needs snap, this will be a fine reason to switch from Ubuntu to Debian.
-> all set the screen to 800×600, even though most installers don't fit into such a low resolution
I've never quite understood why they do this. It's such an easy thing to program for and/or to test. I've experienced something like it but I can't remember what it was now. The tops of the buttons were just about visible and tabbing to the button I wanted highlighted it.
-> Finally, at 19GB, it deigned to try.
That really is bloat. I wonder where it is all going. It's probably worthy of an article on its own - the bloat of Linux today. How much more is one getting in this release compared with a release from one or two years ago?
-> The installer boots using X11, but once installed it defaults to Wayland.
This is what gets me about some of these distros. Why have both? If Wayland is not considered good enough during the install then attention should be given to make it good enough. Now we see where some of the bloat is.
-> rclone
A very good tool.
"I wonder where it is all going. It's probably worthy of an article on its own - the bloat of Linux today."
Boy, can I tell you some stories.
Here's one place it's going:
https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/tree/
you see that gigantic pile of iwlwifi-(blah,blah) files? Intel just loves churning out increasingly ridiculous piles of binary firmwares for its wifi adapters. We have to ship all of them or else your wifi won't work.
I have been told that AMD or NVIDIA or both (I forget) is planning to start doing the same with its graphics cards, only in that case each file won't be ~1MB, they'll be ~20MB. That's going to be fun.
There's a whole boatload of fun to be had once you start trying to unpick "bloat". You can read up on some of it at https://bugzilla.redhat.com/show_bug.cgi?id=2149246 . See the top 20 packages by size on a Fedora Workstation live image: https://bugzilla.redhat.com/show_bug.cgi?id=2149246#c62 . Number one is libreoffice, which is 300MB (uncompressed) on its own. Firefox is 233MB. Distros can't really do much about that except not ship them, which probably wouldn't be too popular. Next is glibc-all-langpacks , which is basically a bunch of translations for core system strings to a zillion languages; you can save that by only installing the languages you want. Next is Java, which is there for LibreOffice. Next is linux-firmware, which is a giant pile of upstream vendor-related pain that we (I) trim obsessively but it's a gradually losing battle. ibus, libpinyin-data and google-noto-sans-cjk-vf-fonts are necessary for seeing and typing CJK (that's Chinese, Japanese, Korean) text, which is, you know, quite important to a lot of folks. llvm-libs is needed by mesa, a core part of the graphics stack - mesa very much wants to be built with llvm/clang, not gcc. cldr-emoji-annotation, well, the kids these days can't live without emojis. mesa-dri-drivers makes the pictures show up on the screen. webkitgtk6.0 and webkit2gtk4.1 are different versions of the same goddamn HTML rendering engine because GNOME hasn't managed to get everything onto the newer version yet (so our live images get Firefox's engine and two versions of webkitgtk. Fun.) Then you've got two bits of the kernel, geolite2-city is...uh, hey! We might be able to get rid of that one. It seems it's only there because ipcalc recommends it. I'll look into that. gnome-user-docs does what it says on the tin, firefox-langpacks is localization for firefox, podman is like docker only better (don't @ me), and python3-libs is kinda important.
so, yeah, we're trying! I am, anyway. sigh.
(still, 19GB does seem like sort of a lot. I haven't checked what the smallest disk you can install Workstation to is lately, but I'm pretty sure it's smaller than that. A Rawhide install I have in a VM here seems to have used up 8.3GB of disk; I know our installer has various pads on the size check, and I bet Ubuntu's does too, so it probably wouldn't let you get away with less than about 10-11GB...)
I would guess part of the problem is the constant move towards "snap" packages? All of the compressed images in /var/lib/snapd/snaps ... all the uncompressed versions in /snap ... plus the cache directories. There are normally two versions of each package (current and previous). It soon adds up.
Re: Finally, at 19GB, it deigned to try.
Bear in mind you are not trying "Linux" there, you are trying full Ubuntu.
Linux itself is small and neat and modular. Ubuntu push their pet projects too hard. Something like Mint is a better starting point if you want to reduce HD footprint but still was an easy start.
Distros can add or remove stuff. I have Ubuntu containers that are just a few megabytes by not even installing systemd.
Most of that 19gb is not a necessary part of the OS, and very little is Linux. It's all good stuff that many people use, or it would not be there, but it's not even an indication of the size of a Linux install.
Why? What value does it have beyond almost but not quite doing what X11 already does? How many people are using a graphical Ubuntu on a multi-user system where the supposed security issues on X11 are relevant?
I don't mind developers wasting their own time but I object to having their half-arsed systems installed by default.
X11 has one benefit, its well tested and works. wayland has tonnes of benefits in terms of its simplicity, architecture, and ease of development, but lacks the bulletproofness of X11. It will get there.
These non-LTS releases of Ubuntu are ideal for testing that out. I would hazard that the next LTS won't be wayland by default, but it will be one day.
This.
And this
"How many people are using a graphical Ubuntu on a multi-user system where the supposed security issues on X11 are relevant?"
Accessible access to computing is likely to be a much more serious security concern for many because I accessible computing can put you life or your money at risk even without a software flaw.
It's a pet peeve that people think locking you out _is_ security. Security is provided by your home and it's comforts.
"The problem with Wayland's simplicity is that it pushes the complex problem of things like accessibility out to the app developer..."
And, unless things have changed, the Wayland team are NEVER going to copy or emulate the X11 client/server functionality. So when I want to run a graphical application remotely, I'll need an entire GUI running on the remote machine and have to use some kind of remote desktop tool instead of just doing an ssh -X $client and run the app I want with no other GUI or overhead running on $client. In effect, downgrading to the Windows RDS model because the X11 client/server model is too hard and they claim hardly anyone uses it.
This.
I've just recently trialled X2Go, which at the moment seems to be the best (free) remote desktop solution. In truth the set-up for ssh tunnelling was not too horrendous - just needed a bit of fiddling on the authentication side. And over 100Mb broadband quality was just about acceptable. But - it meant having to set up at least a minimal desktop environment on a headless server (I went for Openbox) - which, apart from anything else, means admin privileges (and pulling in loads of dependencies which have no place on a server).
The client/server model serves a purpose, and fulfils a need which has not gone away.
> X11 has one benefit, its well tested and works.
It has another benefit which is hugely important to some users: X11 forwarding over ssh. The alternatives under Wayland are a faff and don't really cut it.
This is a show-stopper for me - I won't be using Wayland any time soon.
I've mixed feelings on the topic of X11 forwarding.
On one hand, getting an X11 connection via a pipe, socket, etc is absolutely fantastic, it works very well, suits almost all purposes. And the really good thing is that the quality of the display is "perfect", provided your X server is well set up with the right fonts, etc.
The other approach to remote GUIs - followed by things like Microsoft's RDP - is to forward the whole frame buffer, like a streamed video. I am predisposed against this approach because the quality of the end result is often inferior (compression artefacts, etc). However, the big advantage is (if done properly) that if you've got high frame rate whole frame changes going on a lot, on a slow connection it will (when all the frame changes have stopped) settle down quickly, instead of having to receive absolutely everything and draw it all.
But that's a comparatively rare need these days; high bandwidth connections are commonplace, and any gamer is likely on Windows, X box or PlayStation anyway.
Given that connections these days are generally pretty good, I'd say there was more value than previously in X11 (or an X11 style approach).
High Bandwidth
Looking to the future, one has to ask where we're headed with things like PCIe, Ethernet, etc. Fabrics likes these are now very expensive to develop; loads have dropped out, things like Serial RapidIO cannot compete any longer because the cost to developer the "next chip" is so high. There is a possibility that, one day, the only interconnect we'll have is Ethernet.
Which then begs an interesting question. If the innards of a computer are essentially going to become just a network of components, with IP addresses, etc, that'd mean that down at the electronic level graphics would be client-server, just like X-11 already is.
Yeees... I work from home (on Linux) a fair amount, on a reasonably fast, but not exactly blistering broadband connection. I need to access a raft of remote computational Linux servers via a viciously-firewalled gateway server which only accepts ssh connections. X11 forwarding over ssh (tunneling via the gateway server) Just Works (tm) with acceptable lagginess and good-quality rendering - not great for graphics-heavy clients, but I don't generally need those. In the past I have trialled various flavours of VNC and NX. They were all nightmarish to configure for ssh+tunneling, graphics were generally awful and laggy, and frequently they Just Didn't Work (tm). Perhaps there's some magic-bullet solution out there I haven't discovered yet, but till that turns up I'll stick with what works.
"...in the background, something was going wrong as the System Problem Detected pop-up kept reappearing, no matter how many times we closed the warning."
Remember that from years ago, like 10.10 days onwards. Never did find out what it was. Never seemed result in any issues on the actual reboot into the fresh install.
"That means next version of the most popular Linux distro is getting close"
Most popular distro?
By what metric is that statement based on?
According to Distrowatch Ubuntu is at number six on their page hit ranking with MX Linux at the top with over twice as many page hits.
Ubuntu might be one of the more notorious distros out there but most popular?
"The DistroWatch Page Hit Ranking statistics are a light-hearted way of measuring interest in Linux distributions and other free operating systems among the visitors of this website. They correlate neither to usage nor to quality and should not be used to measure the market share of distributions. They simply show the number of times a distribution page on DistroWatch was accessed each day, nothing more."
As far as I know, Ubuntu has been number one in usage for quite a few years, though I'm not sure where those numbers can be found.
...and you gave it 19GB? I'm not surprised it locked up after a bit, it does need some additional elbow room for temporary files when installing. Even a significant upgrade will fetch up to a couple of GB of files, and unpack them. Look at a df -h after an upgrade and again once you've done an apt clean.
So one of the reasons you are saying that Ubuntu is glitchy because it refused to install on a tiny virtual hard disk that was only marginally bigger than the base OS needed?
Sounds like a DFU problem to me.
...because of course disk space is such a problem these days, I must remember to upgrade my disk to 32Gb when it comes to payday.... <slaps head>
Having started life with HP-UX, TWM and FVWM I’ll comfortably use almost any Linux distrib and DE that’s fit for purpose and I’m the sole grumpy old man here who enjoys SystemD and use it a lot. I default to distributions with APT though for my own machines, usually KUbuntu for the desktop. It works well and almost anything has a package, I have experienced the same can’t be said of Pacman, and Arch distribs also often break stuff in corner cases nobody notices, until you need it.
Am tempted by Pop!_OS though, might spin it up in a VM one of these days. I see it has flatpack instead of Snap as well.
"Having started life with HP-UX, TWM and FVWM I’ll comfortably use almost any Linux distrib and DE that’s fit for purpose and I’m the sole grumpy old man here who enjoys SystemD and use it a lot."
You're not ;) And I, too, started with HP-UX (as well as AIX, Solaris, IRIX, Tru64, DYNIX/ptx, SPIX, ReliantUNIX and probably some others I no longer remember).
"I default to distributions with APT though for my own machines, usually KUbuntu for the desktop. "
I rather stick with the modern version of the mentioned old-time UNIX variants, which is enterprise grade Linux. That means Red Hat (RHEL/Alma Linux/Rocky Linux/Oracle Linux) and Suse (SEL/openSUSE), although on the desktop that's almost always SUSE.
I don't like Ubuntu, not just because there's always something that's broken (it's called "Windows amongst Linuxes" for a reason), but also because Canonical is mostly a taker rather than an upstream contributor (I believe even Microsoft contributes more to Linux than Canonical, while Red Hat and SUSE are both major contributors to Linux and a wide range of FOSS projects).
> while Red Hat and SUSE are both major contributors to Linux and a wide range of FOSS projects).
... which is not always a good thing, in the case of Red Hat, at least. Ref: poettering, systemD, etc.
OTOH they eventually "contributed" him to Microsoft, maybe it'll work out somehow. Though damage done in any case.
i've been using ubuntu for many years but the upgrade from 18 to 22 proved painful. i use it on 2 different lenovo pcs. upgraded the first ok. the second a disaster. a number of issues inc refusal to suspend from all the options. i have to shut down and restart after each session unless i leave it plugged in and permanently on. so i'm now wary about new releases
... next upgrade, just nuke and pave with Debian testing?
Canonical isn't really any worse than Red Hat in pushing their "innovation du jour", it's just that they've become less and less useful. Like Microsoft, they seem to be caught in a vicious cycle of pushing features in search of a use case. I know some people use that stuff, and by all means they should continue. But for Pete's sake, can we just not make them system dependencies?
There's a growing list of extras I have to remove or defang after each install or upgrade. Snaps (and the various tricks to keep them from compromising performance) are really only the tip of the iceberg. Tried lxc, then lxd, and podman: then wound up back on docker (k8s would be overkill for my home use case). After experimenting with alternative solutions from pyenv to AppImage, I'm now mostly back on deb packages, heavily favoring official sources but using PPAs when it makes sense.
Finally, netplan. No. Just no. Using YAML for network configuration is not worth the effort. Neither is cloud init for general purpose install. Put that stuff in a separate "for the cloud" install disk and default us to NetManager (with the option to use the old reliable ifupdown if preferred).
Like I said: Debian testing
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
