Distro developers began discussing ways to reduce the size of firmware updates last year. Now, in Ubuntu 26.04, it’s introducing meta-packaging to spread Linux firmware across 17 smaller packages in the
resolutearchives. This resolves a bug filed in 2022.The sub-packages are:
- linux-firmware-mellanox-spectrum
- linux-firmware-intel-wireless
- linux-firmware-intel-graphics
- linux-firmware-amd-graphics
- linux-firmware-nvidia-graphics
- linux-firmware-intel-misc
- linux-firmware-broadcom-wireless
- linux-firmware-netronome
- linux-firmware-misc
- linux-firmware-qlogic
- linux-firmware-marvell-wireless
- linux-firmware-mediatek
- linux-firmware-marvell-prestera
- linux-firmware-realtek
- linux-firmware-qualcomm-wireless
- linux-firmware-qualcomm-graphics
- linux-firmware-qualcomm-misc
What the hell is this? Ubuntu can’t just go around making decisions I actually agree with!
Cynically, isn’t this just because Debian did it with Trixie, so now Ubuntu’s next version is pulling in the change?
They do it all the time, but then ‘balance’ it with something terrible. (these aren’t in chronological order)
Upstart - good idea.
PulseAudio wayyyy too early - bad idea.
Unity - good idea
Mir (display server) - bad idea
Snap - fuck you
I wanted to create a caching snap proxy and it turns out you have to register it with canonical to get a cert.
I’d thoroughly erased it’s existence from my mind it seems. It’s the reason I went back upstream to Debian many moons ago.
It’s existence alone didn’t bother me, but the day I went to install something with APT and it force installed the Snap was the last day I ever used Ubuntu.
Mint doesn’t use snap, officially doesn’t support it (though it can be enabled and used).
Are there any plans to maintain Unity? I’m seriously bummed about the state of this awesome desktop / window env 😕
I keep running it, but I’m worried one day it will just go away.
Have you ever tried to use Upstart? It was afwul, in practice it was worse than sysvinit+lsb, in a time one woukd thought any new init system can be better.
There was no way to properly define any complex servixe dependencies, especially with optional or alternative components. And making mistake in defining service forking behaviour would open lock the system down so it could not be cleanly shut down. Those were serious flaws in both design an implementation.
I made a mistake trying to use it in a Linux distribution I was co-developing. So much time an effort lost, when we could directly switch to systemd. But systemd was described as ‘work in progress’ an Upstart ‘practically production ready’ then.