Writing this from my Debian system, it's a great distro that has been excellent to me as a daily driver. I switched to Debian 6 after Ubuntu went way downhill and haven't had cause to regret it.
I like Debian's measured pragmatism with ideology, how it's a distro of free software by default but it also makes it easy to install non-free software or firmware blobs. I like Debian's package guidelines, I like dpkg, I like the Debian documentation even if Arch remains the best on that front. I like the stable/testing package streams, which make it easy to choose old but rock-stable vs just a bit old and almost as stable.
And one of the best parts is, I've never had a Debian system break without it being my fault in some way. Every case I've had of Debian being outright unbootable or having other serious problems, it's been due to me trying to add things from third-party repositories, or messing up the configuration or something else, but not a fault of the Debian system itself.
>I've never had a Debian system break without it being my fault in some way.
Debian is great but I can't say this is a shared experience. In particular, I've been bitten by Debian's heavy patching of kernel in Debian stable (specifically, backport regressions in the fast-moving DRM subsystem leading to hard-to-debug crashes), despite Debian releases technically having the "same" kernel for a duration of a release. In contrast, Ubuntu just uses newer kernels and -hwe avoids a lot of patch friction. So I still use Debian VMs but Ubuntu on bare metal. I haven't tried kernel from debian-backports repos though.
> Debian's heavy patching of kernel in Debian stable
Needs citation.
Debian stable uses upstream LTS kernels and I'm not aware of any heavy patching they do on top of that.
Upstream -stable trees are very relaxed in patches they accept and unfortunately they don't get serious testing before being released either (you can see there's a new release in every -stable tree like every week), so that's probably what you've been bit by.
LTS has had major breaking changes in various areas in recent times too, virtio was badly broken at one point this year, as was a commonly used netlink interface. Hat tip to the Arch kernel contributors who helped track this down and chase upstream, as we had mutually affected users. The debian and ubuntu bug trackers were a wasteland of silence and user contributions throughout the situation, and frustratingly continued to be so as AWS, GCP and others copied their kernel patch trees and blindly shipped the same problems to users and refused to respond to bugs and emails.
You're right stability comes from testing, not enough testing happens around Linux period, regardless of which branch is being discussed.
It's not easy testing kernels, but the bar is pretty low.
One of the unsung praises of Arch is that it's turned thousands of users into testers. Before someone says "that shouldn't be the user's responsibility" I'm going to say I'm not so sure. We're all in this together. I'd rather deal with a bug or two on my desktop at home if it means it gets fixed before appearing in a distro that gets used for servers at work and causes issues there where the consequences are much higher.
> One of the unsung praises of Arch is that it's turned thousands of users into testers.
You can do that well enough with Debian's "testing" and "unstable" release channels. Aside from the few months leading up to a new "stable" release, which usually isn't a big deal (and fixing regressions in "stable" should then be a higher priority anyway). Just don't install it on systems that you actually depend on to keep working. But running it on your desktop at home that you only use to play and experiment with is just fine.
I have a similar experience. My not-so-tech-savvy brother also has the same laptop setup I do (arch+XFCE). He knows to yay -Syyu and it's usually never a problem. The recent upgrade there was the vlc package split problem so I told him to hold on upgrading and that I'd come and do it. While I needed to sit and filter and install the optional dependencies myself for my upgrade, a week later it was already figured out (based on user feedback I assume) and the usual yay -Syyu installed just the right optional dependencies.
I don't consider myself particularly adept with linux. I've only been running it daily on the desktop for the last few years and, aside from mucking around with TWMs, I've not done much poking about with the internals.
Despite the reputations, I've had far fewer issues on Arch-based desktop distros than back when I was rolling Ubuntu and Debian.
Yeah same. I think the release cycle actually doesn't matter at all. The reason for it is that the majority of breakage are caused by components/extensions of gnome and kde and non-DE-yet-complex software in distros with a lot of those present out of the box, like manjaro, breaking backwards compatibility every other week.
When people switch to arch they typically set things up from scratch, end up choosing simple tools and avoid most of the unstable stuff distros push onto you.
The virtio bug bit me. I have one hostnode (debian) with one nic that gets passed through to a virtualized opnsense. Storage is two consumer nvme in raid0 as system disk. I was expecting downtime with this setup, but kernel bug was not on my list.
The folks behind Debian LTS and Freexian ELTS are all Debian members/contributors, and the Debian LTS changes end up in the Debian archive, while the Freexian ELTS ones are publicly available, just in an external archive.
IMO, considering the size and scale of the kernel (millions of lines of code, variety of architectures supported, # of subsystems and ridiculous amount of device drivers ), these patches might as well be counted as nothing. I'd say they're basically shipping a pristine kernel :D
These days all of my “Debian” bare metal systems are technically running Proxmox, which I think is a relatively happy medium as far as the base Debian system goes — the Proxmox kernel is basically the Ubuntu kernel, but otherwise it’s a pretty standard Debian system.
I’ve thought about (ab)using a Proxmox repository on an otherwise stock Debian system before just for the kernel…
Same at $work all physical servers run proxmox VE (by policy), 90% of VMs are debian (cloudinit genericcloud), the rest misc linux and various windows.
The upstream kernel already backports enough regressions on its own to its stable releases, Debian's kernel team does not help them too much with that.
Integrated Intel GPU and no graphical system, just KMS VT (text console). That's what made it so frustrating - only displaying a console should not result in kernel panics under CPU load! Admittedly, the experience was anecdotal and years ago and I heard Debian is doing less of a RHEL-style "frankenkernel" now.
drm/i915 was a pretty miserable experience for me on one machine. The Intel drivers for that chipset around the 5.3 kernel era weren't good, I recall lots of bug reports at the time. Below is one of the several issues that I was affected by
Intel's integrated GPU driver team, actually all driver teams, had a period of frequent screw-ups a while back (five years ago? Time flies). They also borked e1000e driver in the same period.
On the other hand, I had and still have many Debian installations, some with Intel integrated graphics. None of them created any problems for a very, very long time. To be honest, I don't remember even any of my Intel iGPU systems crashed.
...and I use Debian for almost two decades, and I have seen tons of GPU problems. I used to write my Xorg.conf files without using man, heh. :)
Yeah, I ditched Ubuntu Server after too many upgrade headaches. I manage 75+ VPS instances for app hosting, and it's nerve-wracking doing maintenance updates knowing there's a chance one won't boot after. That's easily an extra 1-2 hours per VPS just to get it back. Switched to Debian back in the 8.x days in 2015 and it's been smooth sailing. Never had it break unless I was the one who messed it up.
Me too. All the server software (postgres, caddy, bun, etc) I'm using runs just fine on Debian, and I never have had updates break something on my Debian servers.
The only thing I can say against Debian is that it tends to start new server software immediately after install, before I have a chance to configure it properly. Defaults are sane for most packages, but, still, it scares me a little. In that I like the Red Hat approach of installing and leaving it off until I decide to turn it on.
Just have sane firewall rules and you are good. E.g. if I install openssh-server and it auto starts, it doesn't make it out of my machine because my nftables does not allow inbound on port 22. It's just knowing the default behaviour and adjusting your practices for it.
No, because you can install and configure the firewall before you install package X. (without knowing anything about X, your firewall defaults can just prevent X from doing anything)
But you can't (easily) configure package X itself before you install it; and after you install it, it runs immediately so you only get to configure it after the first run.
> And one of the best parts is, I've never had a Debian system break without it being my fault in some way. Every case I've had of Debian being outright unbootable or having other serious problems, it's been due to me trying to add things from third-party repositories, or messing up the configuration or something else, but not a fault of the Debian system itself.
You're not trying hard enough ;-)
I have Debian on an old MacBook Pro and had it on an even older iMac, and I've had a few problems over the years. Always with proprietary drivers - WiFi, graphics, webcams, etc. - Apple really don't want people using free software on their hardware. There's always been a fix, but there have been a few stressful moments and hoops to jump through.
But it's definitely my favorite distro, and I run it everywhere I can. Pretty much always "just works" anywhere but Apple.
For me, it was a combination of Ubuntu breaking upstream and introducing its own unnecessary systems.
I had a few issues caused by Ubuntu that weren't upstream. One was Tracker somehow eating up lots of CPU power and slowing the system down. Another was with input methods, I need to type in a pretty rare language and that was just broken on Ubuntu one day. Not upstream.
The bigger problem was Ubuntu adding stuff before it was ready. The Unity desktop, which is now fine, was initially missing lots of basic features and wasn't a good experience. Then there was the short-lived but pretty disastrous attempt to replace Xorg with Mir.
My non-tech parents are still on Ubuntu, have been for some twenty years, and it's mostly fine there. I wouldn't recommend it if you know your way around a Linux system but for non-tech, Ubuntu works well. Still, just a few months ago I was astonished by another Ubuntu change. My mom's most important program is Thunderbird, with her long-running email archive. The Thunderbird profile has effortlessly moved across several PCs as it's just a copy of the folder. Suddenly, Ubuntu migrated to the snap version of Thunderbird, so after a software update she found herself with a new version and an empty profile. Because of course the new profile is somewhere under ~/snap and the update didn't in any way try to link to the old profile.
Then there were stupid things like Amazon search results in the Unity dash search when looking for your files or programs. Nah. Ubuntu isn't terrible by any means but for a number of years now, I'd recommend Linux Mint as the friendly Debian derivative.
As I understand it, snap the package format is not proprietary. Its as open source as say flatpak. What is proprietary is Canonical official snap store, and they patch their version of snap to only use that store. It'd be the same as flatpak being tied to only flathub.
Of course that goes against the spirit of FOSS, but there's a bit more nuance there than simply saying "snaps are proprietary".
Snaps don't just suck from an ideological but also practical perspective, as described for Thunderbird. Firefox on Ubuntu has also serious permission issues with webcam support OOTB even experts are struggling with (involving AppArmor, pipewire, snap, and FF device config). and has become unusable for things like browser-only MS Teams on mainstream notebooks.
Containers, popular as they may be on servers, can only add breakage and overhead to desktops, especially for an established and already much better organized system like Debian's apt. There just haven't been any new desktop apps for way over a decade that would warrant yet another level of indirection.
I've tried making snap packages, but I discovered they're very tightly tied to Ubuntu's base packages. They're not portable at all. In essence they're effectively just a secondary Ubuntu-specific package format for user-level applications.
For example, with flatpak you select a base runtime for your package that contains mostly system-agnostic libraries. With snap, you specify an Ubuntu version as a base runtime and additional dependencies that are Ubuntu packages.
My understanding is that the base layer (similar to what FlatPak provides) is shared and downloaded by the snap manager so it is portable as long as you want to download it.
The end result should be similar to FlatPak where you have practically no dependencies as it should package almost everything.
Snapd still hardcodes Canonical's snap store signing key and provides no mechanism to add your own keys. Any other snap repos will be treated as second class citizens.
Defaults matter a lot, and snaps are the default in Ubuntu.
The topic is not whether snaps are avoidable or not, but the Ubuntu is going downhill. And snaps are purported to be part of that downhill, which would be Ubuntu's NIH syndrome. As far as I know, Ubuntu's only successful development is Ubuntu itself - the other projects have all failed over the years, and snap, while ongoing, is not winning any popularity contests either.
Snaps per se are no better or worse than flatpak. Canonical's mistake, IMO, was to make their store the only place snaps can be hosted. That is the "proprietary" bit everyone keeps talking about.
But in practice even for flatpak the only realistic place you can publish your flatpak if you want any traction at all would be flathub, so both formats have only one store right now. But flatpak allows a custom store while for some strange reason Canonical decided not to allow snap that freedom.
Another problem is, Canonical promised to release server components and enable alternative stores, and just forgot that they made that pledge.
Also, rugpulling users and migrating things to snaps without asking their users in order to "create a positive pressure on snap team to keep their quality high" didn't sit well with the users.
> But in practice even for flatpak the only realistic place you can publish your flatpak if you want any traction at all would be flathub
But, for any size of fleet from homelab to an enterprise client farm, I can host my local flathub and install my personal special-purpose flatpaks without paying anyone and thinking whether my packages will be there next morning.
Freedom matters, esp. it that's the norm in that ecosystem.
I was neutral-ish about Ubuntu, but I flat out avoid them now, and migrate any remaining Ubuntu server to Debian in shortest way possible.
I'm using Debian for the last 20 years or so, BTW.
Yes, same. I started with Ubuntu back in the day, because the server I inherited ran Ubuntu, and it was just natural after that for me to run it on the desktop as well. I grew to dislike their NIH over the years, tried distro hopping, and settled on Debian.
Yes, I agree. Snaps or Flatpak, not much of a practical, technological difference. What sets them apart is the way the distribution is handled, including the open source availability of the backend, which enabled for example Red Hat and Elementary to run their own stores.
If you are making your own distro, creating your own flatpak store is trivial, that's all what matters. Linux Mint doesn't use snap exactly because Canonical forces everyone to use their snap store.
Canonical doesn't force anyone to use anything. Snap is open source, just modify it to use a different store if you want. Mint literally forked a zombie DE, but changing a few lines of code in snap is an issue...
Defaults matter a lot, snap is not open source (client is, backend isn't), you cannot "just modify it (Ubuntu)" to use a different store, because Ubuntu installs snaps even with apt. Mint is not part of the discussion.
Red Hat do the same. They reinvented the wheel on multiple occasions (systemd and it's whole ecosystem like systemd-resolved and timed and the whole kitchen sink; podman, buildah, dnf, etc etc.)
They just have more success on getting their NIH babies accepted as the standard by everyone else. Canonical just fail at that (often for good reasons, Unity was downright crap for some time) and abandon stuff, which doesn't help their future causes.
Canonical did their own NIH init daemon called Upstart which failed due to the fundamental design and the implementation being plain bad. Redhat builds better software which is why their NIH gets more adoption.
They're not forced on anybody, they're not required by systemd, and many distributions use more feature-rich alternatives (including, afaik, RHEL — last time I looked at it, they used dnsmasq and chrony). They're also often shipped as separate optional packages:
> Still not anywhere near as popular as Docker. Although technically they're far better than Docker, and if anyone is using them, it's for that reason.
NIH packages are generally expected to be less popular, yes. They have some technical merit, though in my opinion that's mostly trade-offs rather than one being strictly better than the other. I would be surprised if everybody using them is using them because of technical merit as opposed to it being pushed by the distro.
Red Hat builds really good stuff. NIH is sometimes right because nobody invented the stuff at all. Standard Unix tools are great but they don't solve everything, so we've ended up with most distros having "the Debian way" or "the Red Hat way", the main difference of course being deb/apt/dpkg vs rpm/yum/dnf. When building an embedded system with Yocto, the basic choices are also Debian or Red Hat style, though you can of course do anything.
Special mention goes to NetworkManager, which has become the de facto standard way to configure networking because it's good. And with nmcli I can even remember how to connect to wifi from single user mode.
>They just have more success on getting their NIH babies accepted as the standard by everyone else.
This depends on the phrasing. We could also say that Red Hat produces actually useful software, in contrast with Canonical, whose developments don't seem to provide value over existing solutions.
We could also say that Canonical tries really hard to do exactly what Red Hat does, but in a slightly different space, and not very successfully.
A major difference is that Canonical projects have copyright assignment policies, while Red Hat projects don't - this probably explains a lot of the difference in adoption dynamics.
You're right. You don't have to use snaps. Ubuntu migrates packages slowly in behalf of you.
Using apt to install some packages installs snap plumbing and downloads the package as a snap automatically. You don't have to install it manually.
There's no malicious intent though, it's made to "impose a positive pressure on the snap team to produce better work and keep their quality high" (paraphrased, but this was the official answer).
Installing the inferior snap packages when you apt get is one of the worst cases of a Linux distro refusing to respect the user's intent that I've experienced.
This has got to be the most user-blind self-imposed preference in a modern operating system outside of Microsoft's BS.
If you're going to use an OSS operating system, the control of what is placed on the system should be inherently with the user. If the developer has a question if a new package should be added or is required, throw a prompt and ask -- with a default to not use application containers and the default packaging system.
And one of these migrations broke my workflow substantially enough that a dist-upgrade turned into a complete system reformat to Debian and cost hours that I couldn’t afford.
I'm with my neighbor comments. How do you use Ubuntu without snaps? The base Ubuntu install already comes with several snaps. Installing random things through apt leads to snaps. I personally do not know how to avoid snaps on Ubuntu.
It's just an empty package that tells the system snap is installed, to stop the broken dependency chains you otherwise get from force uninstalling snap.
It's been working fine on a handful of Ubuntu 24.04 systems I've been handed and can't change the OS of, for about half a year now.
I forget the exact context but I recall an Ubuntu dev stating they have more users of the Firefox snap alone than trendy distros have entire users.
I think it’s worth keeping that in mind with all the hate Ubuntu gets. Most users are just silently getting their work done on an LTS they update every two years.
Well, I don't know which trendy distro the dev is referring, but Debian is complete opposite of Trendy. It's a bedrock distro silently running almost everywhere in some form or shape.
Most of the Linux-based (enterprise and/or embedded) appliances are built upon Debian, for example.
P.S.: The total number of Debian installations and their derivatives are unknown BTW. Debian installations and infra do not collect such information. You can install "popularity-contest", but the question defaults to "no" during installation, so most people do not send in package selection lists, unlike Canonical's tracking of snap installations.
LXD was forked as Incus, and it’s an absolute delight.
Seamless LXC and virtual machine management with clustering, a clean API, YAML templates and a built-in load balancer, it's like Kubernetes for stateful workloads.
Incus is fantastic. I think Proxmox is where everyone is migrating to after the VMWare/Broadcom fiasco, but people should seriously consider Incus as well.
Switching to rust-based system software is very different from the clearly profit seeking (or control seeking which is just long term profit seeking) changes like ads and snap (with massive friction to not using snap).
Yes, but I prefer glibc + GNU Coreutils based systems in my installations. They're additional nails on top of the (fatal) ones like snap, Ubuntu Pro and MOTD ads.
> in what way has it gone uphill versus just using Debian?
Their lawyers' willingness to risk shipping pre-built zfs kernel modules (that are always in sync with the kernel). Pretty important if you're into that sort of thing, it's easier to remove cruft once post-install than to keep an eye on DKMS for years (making sure that it hasn't disassembled itself and continues working).
I'm an old-school user so I'm not exactly Ubuntu's target audience, but for Ubuntu was bad about a lot of the older, lesser-used bits of Linux.
The two things I can remember were problems with NFS out of the box (outside having to install nfs-common, which I'm fine with) and apt-cache not displaying descriptions of packages. There were lots of other, minor annoyances that affected people like me but wouldn't affect someone who got into Linux desktops after, say, 2010. My memory sucks though so those are the two I remember. Yes, there were bug reports filed and yes, they sat in the tracker for years with no attention from Ubuntu.
I wound up back on Debian once I got old enough that I didn't care about being behind the times a couple years.
> I like dpkg, I like the Debian documentation even if Arch remains the best on that front.
That's curious, because when I was learning to make Debian packages, I found the official documentation to be far better than I had seen from any other distro. The Policy Manual in particular is very detailed, continually improving, and even documents incremental changes from each version to the next. (That last bit makes it easy for package maintainers to keep up with current best practices.)
Does Arch have something better in this department?
Are you perhaps comparing the Arch wiki to Debian's wiki? On that front I would agree with you.
> I've never had a Debian system break without it being my fault in some way.
My experience has been contrary to that. I'm a Linux user of 25+ years
with various distros but about half of that time with Debian as my
main desktop. I broke up with Debian about ten years ago thinking we
could still be friends, but every time I've tried to put it on a new
box it since then something weird has happened, most recently about a
month ago on a completely new Intel N150, when it gave me some stick
about video modes. Today my laptop got hosed by an attempted upgrade
from bookworm to trixie, as in tons of error messages and then no more
docker and no more virtualbox. No harm done because Debian taught me
long ago to store a copy of the whole root filesystem on external
media before an upgrade, but now the clock is ticking until I have to
migrate off it or get stuck with something too old to be compatible
with anything.
> I like Debian's measured pragmatism with ideology
There is plenty that could be said of Debian but as far as I’m concerned that’s not part of it.
Debian patches software for purely ideological reasons because they think they are not free enough. That’s not pragmatism. That’s the reverse of pragmatism. It certainly is a real drag on the teams developing the software they try to ship.
You weren't around for when they broke the OpenSSL random number generator for no good reason. That was back in 2008 and it created vulnerabilities that persist to this day. https://16years.secvuln.info/
I still use Debian but it's hard to forget stuff like that even after all these years.
Though I agree with you, if that's the bar you're setting then Debian comes out far ahead of any other OS that I've ever used - Linux based or not. I can recall dozens of worse Windows bugs, most of which did not even affect me because I was not using Windows at the time. Mac has its share too.
I think this is all true, but the "being my fault" part has gotten better for me with nixos. Broke it? just reboot into the previous version and get configuration.nix back from git. I had to reinstall exactly once in 2016 shortly after the first install, but I don't know what I did wrong. the third time I installed nixos was last week when I bought a new computer that came with Windows.
I've been always a Fedora person, still am. But my PC moved to Proxmox (debian) in 2023. Now a Fedora Atomic sits in a VM running flatpaks and podman containers :D
You don't mention say what you like specifically about Debian, most of what you wrote could be said for a lot of distributions.
So here is what I _don't_ like about Debian :-)
- I don't like Debian package tooling (dpkg, debootstrap, de build...). Actually I hate everything about the experience of Debian packaging. Every time I package for Debian, I end up with a messed up setup of chroots and have to make triple sure nothing leaked from my environment.
- Debian has a habit of repackaging everything at their own sauce, disregarding upstream philosophy. Debian packages will have their own microcosm of configuration directories, defaults, paths, etc. orthogonal to what a pristine installation look like.
- Debian has the annoying habit of default starting installed services. So you always have to dance around your configuration management to disable services, install them, configure them, then restart them.
Thank you to all the Debian volunteers that make Debian and all its derivatives possible. It's remarkable how many people and businesses have been enabled by your work. Thank you!
On a personal note, Trixie is very exciting for me because my side project, ntfy [1], was packaged [2] and is now included in Trixie. I only learned about the fact that it was included very late in cycle when the package maintainer asked for license clarifications. As a result the Debian-ized version of ntfy doesn't contain a web app (which is a reaaal bummer), and has a few things "patched out" (which is fine). I approached the maintainer and just recently added build tags [3] to make it easier to remove Stripe, Firebase and WebPush, so that the next Debian-ized version will not have to contain (so many) awkward patches.
As an "upstream maintainer", I must say it isn't obvious at all why the web app wasn't included. It was clearly removed on purpose [4], but I don't really know what to do to get it into the next Debian release. Doing an "apt install ntfy" is going to be quite disappointing for most if the web app doesn't work. Any help or guidance is very welcome!
> The webapp is a nodejs app that requires packages that are not currently in debian.
Since vendoring dependencies inside packages is frowned upon in Debian, the maintainer would have needed to add those packages themselves and maintain them. My guess is that they didn't want to take on that effort.
Nodejs itself is, but when you install a node project manually, you type npm install and wait while it downloads the 500 different packages it depends on.
Debian follows the same philosophy as for other more traditional languages and expects that all these dependencies are packaged as individual Debian packages.
Just jumping in to say that this is making me genuinely reconsider adopting a licence/policy that forbids repackaging: the fact that someone can repackage my project, but worse, and still use my project's name? Absolutely not. I do not want the burden that inevitably comes when people complain to me that this or that is missing from a repackage.
I mean, that's just how OSS works. Anyone can fork your thing, do whatever, and call it a day. Going MIT or whatever won't save you either - this repackaging business is basically the entire business model of AWS.
Forking is fine, do whatever, but as soon as you make actual changes to the code then adopt your own name. This is what trademarks are for, it's just that official trademark registration is somewhat inaccessible (eg: cost) for open source projects. Could you imagine trademarking every little project you make just in-case it gets repackaged by someone who tears huge chunks of it out?
Debian sources need to be sufficient to build. So for npm projects, you usually have a debian-specific package.json where each npm dependency (transitively, including devDependencies needed for build) needs to either be replaced with its equivalent debian package (which may also need to be ported), vendored (usually less ideal, especially for third-party code), or removed. Oh, and enjoy aligning versions for all of that. That's doable but non-trivial work with such a sizable lockfile. If I would guess the maintainer couldn't justify the extra effort and taking on combing through all those packages.
I also think in either case the Debian way would probably be to split it out as a complementary ntfy-web package.
> As a result the Debian-ized version of ntfy doesn't contain a web app (which is a reaaal bummer), and has a few things "patched out" (which is fine).
My advise to you is to deny all support from people using the Debian version of your software and automatically close all bug tickets from Debian saying you don’t support externally patched software.
You would be far from the first to do so and it’s a completely rational and sane decision. You don’t have to engage with the insanity that Debian own policies force on its maintainers and users.
I agree that maintainers should not be expected to support patched versions of their software, but as a user I like the Debian policies you call insane. I would actually pick Debian exactly because they are cautious with the dependencies.
Debian is not cautious with the dependencies. Debian breaks a lot of what they ship, sometimes flagrantly like removing a whole feature, sometimes insiduously by introducing new bugs. I don't really care that Debian doesn't view it as breaking things. From my point of view, users trying to get my product get subpar experience in a way which is far from explicit.
I personally wouldn't use Debian but people are free to do whatever they want. I don't want to waste my time dealing with Debian maintainers and how they think software should work however. I advise all software developers to do the same and am vocal about it because it's easy to get guilt tripped in the idea that you should somehow support their users because they want to use your product or that introducing changes to support their esoteric targets somehow make sense because they have done the work despite the burden of futur support actually landing on you.
I want to make clear to people who decide they have no interest in it that they are not alone and it's perfectly fine.
And to be clear, I am singling Debian here because they are by far the worst offender when it comes to patching but the comment applies equaly to any distributions that apply invasive patches.
Debian IS more cautious with dependencies, in that you won't get hidden dependencies that aren't in the repos.
I don't want to install an app that downloads and executes 500 node packages that I don't know what they do. Those packages should already be vetted and in Debian. If not, then I'm not interested.
Side stepping the distro repos for dependencies for software in the repos leads to unexpected behavior.
> Debian IS more cautious with dependencies, in that you won't get hidden dependencies that aren't in the repos.
For a definition of cautious I don't personally share.
Debian doesn't vet packages. Debian maintainers are less competent than the "upstream" they question approximately all the time, which is why they keep breaking stuff in more or less severe way (OpenSLL anyone?). And let's not even talk about the insane stuff like when maitainers decide to support a fork they like instead of the piece of software users actually want (Libav anyone?).
> If not, then I'm not interested.
And that's your choice. That doesn't mean developers should care, nor that it is actually a good idea.
Debian has been the stable footing of my Free computing life for three decades. Everything about their approach — from showing me Condorcet, organising stable chaos, moving forward by measured consensus, and basing everything on hard wrought principles — has had an effect on me in some way, from technical to social and back again.
I love this project and the immeasurable impact it has had on the world through their releases and culture.
> i386 is no longer supported as a regular architecture: there is no official kernel and no Debian installer for i386 systems. The i386 architecture is now only intended to be used on a 64-bit (amd64) CPU. Users running i386 systems should not upgrade to trixie. Instead, Debian recommends either reinstalling them as amd64, where possible, or retiring the hardware.
Impressive that i386 support made it all the way to August 2025. I have Debian 10 Buster running on a Pentium 3 which only EOL'd last year in June 2024. It's still useful on that hardware and I'm grateful support continued as long as it did!
OpenBSD still supports i386 for those looking for a modern OS on old 32-bit hardware.
Hopefully i386 (or perhaps a new i386-like port with added support for 64-bit time values) can move to the unofficial Debian Ports infrastructure for Debian 14 (forky) or Debian 15 (duke). Debian Ports has a m68k port, so supporting one for i386 shouldn't be a huge problem.
To what end? Outside of sheer nostalgia if you are running ancient hardware, you probably have a bespoke application which requires that environment. Either you cannot change for hard technical, compliance, or just fear of the unknown. Firewall it from the internet and continue to run whatever release last worked.
I am not happy about unnecessary ewaste, but an i386 almost certainly has and order of magnitude less horsepower than a raspberry pi or N100.
Debian's tagline is the "universal operating system". It's a distribution with active ports on a very large number of architectures [1], even incredibly obscure ones.
The goal of universal compatibility that separates the Debian project from commercial software and even other open-source projects.
The legacy x86 architecture is still far more popular than some that platforms that Debian advertises as having official support for and there has been x86 based processors manufactured for niche applications until recently, eg, AMD Geode and others.
I find it really unfortunate Debian Project is removing official support for new x86 installations. The silver lining is it seems like they'll be an unofficial port and it's likely niche distributions like MX Linux and AntiX will maintain their own builds.
It would be ideal if open-source can develop stronger mechanims to keep support for the large numbers of these relatively niche architectures (eg, through increased usage of emulation over real hardware).
My Linux machine is very modern, but I still need i386 architecture support installed, because Steam requires 32-bit support. And Steam requires 32-bit support so people can play 15-year-old games.
(Admittedly, the 32-bit support Ubuntu ships is less than a full OS and you can't install Ubuntu on a 32-bit machine these days)
Debian is doing basically the same thing Ubuntu is with regards to i386. Packages are still being built for the architecture, but i386 systems aren't supported, and there's no 32-bit kernel package.
According to Passmark the Pentinum 4 1.3Ghz is 55 times slower than a Raspberry Pi 5, so I'd guess it's at least two orders of magnitude. The original Pi is 16 times faster than a P4 1.3Ghz...
You can recycle e-waste (and yes, I know SOME e-waste ends up in China/India/etc. Not all does.)
The e-waste is of substantially less concern than the massive difference in carbon footprint from power consumption.
The i386 architecture hasn't been dropped, it is still available in the archives to support 32-bit applications. The major change is that there no longer is a 32-bit kernel in the archive (the package linux-image-686 is no more). But most packages are still available in their i386 versions:
If the official i386 arch was built with instructions that your hardware doesn't support, tough cookies
While theoretically possible, that would only happen on processors older than 30 years. Debian's i386 architecture still uses -march=i686 as its baseline compiler target, which is the venerable Pentium Pro: https://en.wikipedia.org/wiki/P6_(microarchitecture)
Not sure why you are downvoted, I guess people don't believe this is true. To confirm: The AMD Geode LX was a <5W 32-bit x86 processor which did not support SSE instructions, and is therefore not fully i686 compatible. According to Wikipedia, it was produced until 2019: https://en.wikipedia.org/wiki/Amd_geode#AMD_Geode
It was used in the OLPC XO-1. The Cisco ASA line of firewalls also used Geode processors at least at some point in its lifetime.
Much of that information has to do with creating a new hardware port from scratch. The i386 support just needs to be "demoted" to the Debian ports infrastructure once it's officially scheduled to get dropped from the main Debian repository (which could well happen starting either in Debian forky or duke), and this can probably be done with some special handling.
(Answering the "to what end?" question, a lot of 32bit-only hardware is still available and dirt cheap in the second-hand market (e.g. early "netbooks"), much of it quite well-built and enjoyable to use. While such hardware can no longer realistically browse the "modern" web, it can still find a lot of use for more lightweight tasks, including acting as a "thin client" for more powerful machines.)
Well, the existing i386 port is going to remain as-is for supporting old software (especially games) (but the CPU baseline will likely get increased), it isn't going to be removed, so you would need a new architecture for 32bit-only hardware.
Since i386 is not going to do the 2038 transition either (since that would break the ABI), also you would need to either make a new ABI for the new port, or do the 2038 transition for it too.
Over time more and more 32-bit bugs will get introduced, so there will be lots of maintenance work to do too.
I had seen conflicting information about this, though nothing official. I suppose we'll know more once some actual release planning happens for forky, which might take some time.
When distros mention i386 support they often actually refer to i586 or i686, yes.
True i386 support would mean compatible with the original Intel 386 processor from 1985. The 486 added a few additional instructions in 1989 but things really changed with the Pentium in 1993 - that gave us i586 which is the bare minimum for most modern software today. Much software can still run on regular Pentiums today if compiled for it, but SSE2 optimizations requires at least a Pentium 4 or Core CPUs instead.
I play with retro PCs often and found OpenBSD's i386 target stopped supporting real 386 CPUs after the 4.1 release, and dropped support for i486 somewhat recently in 6.8. It now requires at least a Pentium class CPU, i586, though the arch is still referred to as i386 likely because it's a common proxy for "32-bit".
Debian 3.0 was the last version that ran on the 80386 processor proper, but even as the CPU requirements for the "i386" architecture moved up to the 486, then Pentium, then Pentium II, the name stuck around. Partly from inertia, partly to not break the entire existing mirror infrastructure.
Yes, 386 and 486 stayed relevant throughout the 90s because the price tag for new shiny processors is always higher, and it was not uncommon for customers to favor more or faster RAM/Disk space/graphic card/sound card (that was a thing back then) over better-looking CPU benchmarks.
The Linux kernel also requires at least i486 now. AIUI that decision had to do with smoothing out multicore/SMP support - which is a bit silly because no real 80386 systems in common use are even SMP, let alone multicore. But anyway.
Linux ran fine on 386 chips - that was actually what it was originally developed on. But Intel added a bunch of functionality in the 486, Pentium, and Pentium Pro chips. At some point the powers that be didn't see any value in continuing to support pre-P6 chips anymore.
It was a bit of a strange decision since there were undoubtedly more 386, 486, and Pentium users than some of the platforms Linux continued to support, but that's the choice they made. But they weren't alone. Even NetBSD requires a 486DX or better.
Freexian is for-profit, and all the LTS/ELTS contributors are Debian maintainers, and LTS is part of Debian, while ELTS is publicly available too, but in an external archive.
Ah, they advertised non-profit at one point, but I see that's changed. That may have been "we seek no profit" not "non-profit entity". Thanks for the info on this point.
Back to LTS:
Debian LTS is not handled by the Debian Security and Release teams, but by a separate group of volunteers and companies interested in making it a success.
To the point, Freexian is 100% not Debian, not "part" of Debian, it merely uses Debian's infra gratis for LTS. This does not detract from the good work they do, but we must also not confuse a private company, and its goals, with Debian and its goals.
LTS tries its best, but only supports what it can. Not its fault. Thus they do give preference to packages which are more widely used, and which they have received donations for.
So wildly popular things such as apache2, mariadb, and so on are very much going to be handled. Some rare package which has 400 users worldwide? Not so much.
LTS will very much take patches and any help, but that still ties in to the number of users. If a packages has 400 users worldwide, and most have moved on to the next release? Well, I hope you see my point.
(I've moved customers off of LTS for using rare packages, whilst reassuring them that LAMP servers are very much supported due to this. Popularity counts here, due to efforts of volunteers and externals.)
--
ELTS only supports a further subset of packages. It's not "full" support. I think one would be exceptionally unwise to use it, for say a desktop. That is, unless they were paying for support and had obtained a list of all packages supported.
"Note that when you request a quote, we send you back a list of packages that are not supported or that have limitations in their support so that you can take an informed decision."
Yes, I know that page has a git repo and so on for some support information.
But my points are; not the full distro is supported, you have to track this yourself, you need to be diligent, and even so you need to be sure you're not running rarer packages.
Once again, I do want to reiterate, these are both excellent programs. They do a good job, they're dedicated, but we must be aware of the limitations here.
An example being the differences between security support for main, non-free, contrib in stable Debian:
As you can see, there is no actual guaranteed security support for contrib and non-free. The reasons are logical, however, users need to be aware of the nuance here.
Just as they need to be aware of the nuance of LTS and ELTS.
For example, all of my server installs have non-free, non-free-firmware and contrib blocked via pinning in preferences.d, with only specific absolutely required packages then allowed back in.
(For example I may allow command line apps, but not anything network connected, and only with a once over of functionality and SUID bits and other such things)
--
Really, I see LTS as a crutch that normal users should never use. I suggest we collectively not encourage Desktop users (for example) to use LTS.
The LTS/Freexian people are all Debian members/contributors, so I would not say "Freexian is 100% not Debian" is correct. Basically, some Debian folks got together and started a company to get funding to do an LTS, and also offer other paid services.
At least for bullseye, the LTS team supposedly support all packages, except for games and a few other packages. Its trivial to find out which packages aren't supported too, just run a command, no need to email anyone.
Agreed on the rest, although do note LTS contributors are paid, the security team probably aren't (although some are).
I think in practice, when contrib/non-free stuff has security updates from upstreams, Debian does get updates in stable/LTS. For example the Intel microcode, or WiFi firmware.
I too feel like Debian having LTS is a waste of time, people should be able to upgrade to the next stable within the one year of regular security support for oldstable.
BTW, Ubuntu security support has a similar issue; main is supported, universe is not.
Yep! I did some testing and found 6.8 was the final version to support 486 CPUs. The reason is the OpenBSD project switched to clang for the compiler in 6.9, and in the process became dependent on Pentium instructions.
However, that doesn't stop one from installing a Pentium Overdrive in an old Socket 3 board and running the latest release. ;)
Hahah yeah, it's just a for fun/hobby PC. It's pretty beefy for what it is - 1.3GHz, 512MB RAM, 512GB SSD on a SATA card, fast AGP card. I can run a Debian desktop at 1080p on here.
There is a weird depends you cannot get around without simultaneously removing and installing in parallel. A Debian bug highlighted the above, with a "-" for systemd-sysv- systemd- as a fix, along with allow remove essential.
After this fix, sysvinit builds with debootstrap were almost identical as to bookworm. This includes for desktops.
As per with bookworm through buster, you'll still need something like this too:
$ cat /etc/apt/preferences.d/systemd
# this is the only systemd package that is required, so we up its priority first...
Package: libsystemd0
Pin: release trixie
Pin-Priority: 700
# exclude the rest
Package: systemd
Pin: release *
Pin-Priority: -1
Package: *systemd*
Pin: release *
Pin-Priority: -1
Package: systemd:i386
Pin: release *
Pin-Priority: -1
Package: systemd:amd64
Pin: release *
Pin-Priority: -1
Yes, MX Linux will be doing this (as a separate ISO, until now they're been able to provide a single ISO that lets you choose between systemd and sysvinit at every boot):
I've been living with sysvinit up until Debian 11. Then it became unusable with lxc containers :(, so I had to bite the bullet. But for the basic system it indeed works.
A new APT sources format "debian.sources" is announced with trixie. The now older "sources.list" format is still supported, but is likely to be deprecated in a future Debian release.
See below:
APT is moving to a different format for configuring where it downloads packages from. The files /etc/apt/sources.list and *.list files in /etc/apt/sources.list.d/ are replaced by files still in that directory but with names ending in .sources, using the new, more readable (deb822 style) format. For details see sources.list(5). Examples of APT configurations in these notes will be given in the new deb822 format.
If your system is using multiple sources files then you will need to ensure they stay consistent.
"apt modernize-sources" command can be used to simulate and replace ".list" files with the new ".sources" format.
Modernizing will replace .list files with the new .sources format, add Signed-By values where they can be determined automatically, and save the old files into .list.bak files.
This command supports the 'signed-by' and 'trusted' options. If you have specified other options inside [] brackets, please transfer them manually to the output files; see sources.list(5) for a mapping.
Oh nifty, I hand converted all mine a couple years back. It would have been nice to have that then (or know about it?). I do really like the new deb822 format, having the gpg key inline is nice. I do hope that once this is out there the folks with custom public apt repos will start giving out .sources files directly. Should be more straightforward than all the apt-key junk one used to have to do (especially when a key rotated).
Same. It took me a little bit to get used to it; my initial snap judgment was “this will be more annoying to create via scripting,” but then Ansible added deb822_repository [0] in 2.15 (shortly before Bookworm was released), and then it was no longer a concern.
Ooh, that's nice. Especially nice that it lets you specify both Suites: and Components: in the same stanza, so you don't have to repeat the rest of the line to add -updates and -backports suites.
They only mentioned it briefly, and not by number, but this release includes 95%+ bit-for-bit reproducibility on AMD64, ARM64, and RISC-V across more than 30,000 packages (92% mean across all architectures).
Could you help me understand why the remaining 5% is not bit-for-bit reproducible? For example... if you download a tar of sources pinned to a version, and you run `./configure` and `make` in some kind of container and it doesn't embed some kind of timestamp... why are 95% reproducible and some aren't? Would like to learn/understand.
Hashtables keyed off the address of objects would be an example.
On multiple runs, malloc gives out different addresses (thanks to threads or security concerns) which means things end up in different slots in the table. Then you iterate through it in memory order and you're seeing objects in non-deterministic order, which you do things with.
Embedding file paths / timestamps / git shas and similar was popular for a while too and unhelpful for reproducible builds.
I see that systemd is still doing this thing where they are trying to strong-arm all Linux distros into arbitrary stuff that someone decided is the only right way to do something:
> 5.2.2. systemd message: System is tainted: unmerged-bin
systemd upstream, since version 256, considers systems having separate /usr/bin and /usr/sbin directories noteworthy. At startup systemd emits a message to record this fact: System is tainted: unmerged-bin. It is recommended to ignore this message. Merging these directories manually is unsupported and will break future upgrades. Further details can be found in bug #1085370.
That said, my knee-jerk is also that this is about strong-arming distros. Which leaves a bad taste in my mouth. I’d be interested to hear other viewpoints though.
Note that this is about a different thing - Debian has also merged its /bin into /usr/bin, but now systemd also wants /usr/bin and /usr/sbin to be merged.
Oh thanks, I totally missed that distinctions. You’re right.
My first thought is to wonder: why would systemd care about this then? Any idea?
My second thought is: with a merged /usr, where /usr/{bin,sbin} are on the same filesystem, what’s the benefit of even having a distinct sbin? My understanding is that sbin has historically been useful to have statically linked binaries that can be made available early during boot without needing various lib dirs mounted. But that need seems obviated by a unified /usr…
> with a merged /usr, where /usr/{bin,sbin} are on the same filesystem, what’s the benefit of even having a distinct sbin?
There really isn't any, the sysmlink being for compat. Early boot is nowadays taken over by an initrd/initramfs. Some say what you say sbin was for were actually the root /bin (early boot) and /usr/bin (elsewhere) but it was all just so arbitrary and no-one could actually agree on what is for what actually.
Some say sbin is for static binaries, others for superuser binaries and either destinction really doesn't make sense nowadays as earlyboot is initrd and a lot of "superuser binaries" actually also are fine to use unprivlidged in certain scenarios (e.g. `mkfs.*` for making disk images) so having them in a separate dir is just arbitrary (e.g. `cat` can also fail if I cat the shadow file).
> Debian maintainer (and upstream dev) is open to an upstream patch to add such an option
Wild interpretation right here.
There are only 2 realistic choices: Leave it as is or patch out the warning message in the Debian package
Debian maintainer is clearly deflecting the responsibility here because everyone knows very well that upstream wouldn't accept such a patch.
As it's already explained in the bug report, since Debian has no plan to do that migration in the near future, aforementioned warning isn't only useless and annoying, it's also potentially harmful, thus the correct action would be to remove it downstream like they did it in xscreensaver (https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=819703#84)
But then they'd face the wrath of Lennart, so the only choice left is ignoring the report.
Why do you think this is an attempt at a persuasion tactic? Taint flags in this context just means something that might be relevant to debugging. Which this condition might be, if people in the future are unfamiliar with a potentially anachronistic split between /usr/bin and /usr/sbin. The debug message isn't there to judge the morality of your configuration. It's actively improving your ability to continue to support both ways, by properly indicating what style of system it is for troubleshooting purposes.
> Why do you think this is an attempt at a persuasion tactic
Because they said so:
> As part of that we sometimes adopt schemes that were previously used by only one of the distributions and push it to a level where it's the default of systemd, trying to gently push everybody towards the same set of basic configuration [1]
This was written 11 years before that debug log was added. Extremely loose connection. Things do actually change over the course of 11 years, believe it or not.
It's their modus operandi and it haven't changed, believe or not.
> it is definitely our intention to gently push the distributions in
the same direction so that they stop supporting deviating solutions for
these things where there's really no point at all in doing so. [1]
> Distros may deviate from this by patching this downstream,
but by shipping this as secure default I do hope to gently push
everybody in the same direction.
> start pushing people gently to define GPT partition type UUIDs for missing archs
> Let's generate a single gcc `#warning` message asking people to define
partition type UUIDs for their architectures if they are missing. [3]
Debian choosing systemd (not that it's a new decision) is the reason I'll be switching my Proxmox to FreeBSD/bhyve (FreeBSD has great ZFS support btw).
Once I get the hypervisor systemd-free (no systemd on FreeBSD), I can then install a minimal distro in a VM mean to do containerization (like, say, the Talos Linux distro for K8s, that only has a few executables and they're all immutable) and then I can run containers that, by design, have something that is precisely not systemd as PID1.
So life is good: there's a systemd-free world at the end of the tunnel.
I downloaded both because I intend to seed, but yes, if you have a fast internet connection then minimal is perfect... but if you are on a crappy third-world connection where it might take all day to get the image, it's nice to have it all in place when you are ready to install.
Yeah, I think I remember that, but each time I’m puzzled and not sure. It feels like they hid that deliberately. While Arch is the opposite, it’s so easy to get the torrent file.
Biggest change for me is /tmp behavior. In Debian 13 /tmp become RAM-disk by default (instead of files on the file system) and uses up to 50% of available ram. But as expected of Debian the release notes included an easy fix to restore normal /tmp behavior for people and applications that place many small or large files there.
>"You can return to /tmp being a regular directory by running systemctl mask tmp.mount as root and rebooting."
I kind of wish the distros had decided on a new /tmpfs (or /tmp/tmpfs, etc) directory for applications to opt-in to using ram-disk rather than replacing /tmp and having to opt-out.
This was discussed a ton in debian-devel. First, the tmpfs doesn't take much space already, and /tmp became a folder where persistence should not be expected over the years.
The problem with /tmp was many people and apps used it as an inter-user communication medium and expected persistency there, so it created both security problems and wasted disk space over time.
Since not many packaged apps used the /tmp like that and used the folder the way it should be used, the change was made.
I'm running Debian testing on one of my systems, and the change created no ill effects whatsoever. Not eating SSD write cycles can be considered a plus, even.
However, as I also noted in the relevant thread, the approach might have a couple of downsides in some scenarios.
What terminal emulators default to infinite backscrolling? Genuinely curious here. I've used xterm with a scroll buffer of 512 lines for three decades now and outside of large builds I can't imagine wanting more than that.
Isn't this the feature of /tmp? I set my default download location in Firefox to /tmp exactly for this reason, so that all the junk gets automatically removed after some time. Also, whenever I need a temporary Python script or test a package, I create a venv under /tmp.
On boot has been the standard for a long time and is still the most common. I am personally surprised to hear that now Debian and some distros do it via various automated ways at time intervals.
I think that superkuh's point is that it is not a systemd thing. Cleaning up /tmp by deleting old files has been around since before systemd was invented. Since before Linux was invented, even.
Outwith the world where only Debian and other Linux distributions exist, though, this isn't really systemd doing bad things to poor old Debian. To a wider world it is Debian and other Linux distributions exceedingly slowly reinventing things from Unix.
AIX had /usr/sbin/skulker from at least as early as version 1.2.
For AT&T System 5, Fielder and Hunter were popularizing a similar utility named rmtrash that is run nightly.
Fielder and Hunter exemplify traditional Unix thinking in their 1986 book on Unix system administration, which includes rmtrash:
> By putting all temporary files in one or two directories, it is easy to clean them all out at regular intervals. For this reason, it's a good idea to encourage users to use /tmp for all files they need only a short while.
skulker is not in the old comp.unix.aix Usenet FAQ document, but given the number of times "That's what skulker does, and you should be running it daily." seems to have been the answer over the years, it probably should have been. (-:
To old hands, this is not a systemd novelty at all. The novelty, to old hands, is the idea of doing this not using a script. I remember the war stories from the 1990s and turn of the 21st century when users did things like include LF in the names of temporary files, and administrators suddenly learned the utility of find -print0 and xargs -0 or find -exec rm {} +. Stéphane Chazelas for one has had a lot to say on the subject over the years.
That said, mtree (from 1989) already existed at the point that systemd-tmpfiles was invented, and already had the idea of working from a specification file with names and owners and permissions and whatnot. It is surprising that no-one ever apparently tried mtree for completely wiping /tmp. The BSDs are using find to this day in the several places with they auto-delete stuff in /tmp including the daily periodic, although they did spot that their own find had a -delete option in 1999. (-: Hell, even I am still using find.
Then I think you just don't have much imagination. I have recovered files in /tmp after turning off a machine by booting it back up in single-user mode and accessing the data before it would be cleared in during bootup. Given that "turning off a machine" can also mean "the machine lost power", I can definitely see why people would be surprised by this change.
It could be the 20+ years of using linux where it was standard for it to only have files there deleted on boot. Temp in this case means it's temporary for that boot. Recent redefinitions are surprising unless you've only come to linux recently or from unix 40 years ago.
I am going to assume you are used to tens of days wipings of /tmp. Imagine suddenly it was decided that the wiping needed to happen every tens of seconds for security reasons and to minimize ram usage. Do you think this would impact your workflows that used /tmp? This is 100% analogous to the current situation. Yes, tmp is ephemeral, but on what timescale?
No, I misunderstood the comment I was replying to, and thought they were assuming they could use /tmp as an un-curated dumping ground that was never deleted.
A total of seven architectures are officially supported for "trixie":
"trixie"
64-bit PC (amd64),
64-bit ARM (arm64),
ARM EABI (armel),
ARMv7 (EABI hard-float ABI, armhf),
64-bit little-endian PowerPC (ppc64el),
64-bit little-endian RISC-V (riscv64),
IBM System z (s390x)
It's good to see RISC-V becoming a first-class citizen, despite the general lack of hardware using it at the moment.
I do wonder, where are PowerPC and IBM System z being used these days? Are there modern Linux systems being deployed with something other than amd64, arm64, and (soon?) riscv64?
Mainframes are still holding on in use cases where a single server having continuous uptime is vital. They're designed to have uptime measured in decades, so even components like the processors and main memory have hot spares available and can be hot-swapped without interrupting the OS or running services. They also have continually running system monitoring and diagnostics at the hardware level (not running as an OS service) that will alert both the owner and IBM if they detect some sort of hardware fault. IBM has supported Linux as a first-class OS option for their mainframes since the early 2000s.
From a developer perspective, s390x is also the last active big-endian architecture (I guess there's SPARC as well, but that's on life support and Oracle doesn't care about anyone running anything but Solaris on it), so it's useful for picking up endianness bugs.
Another interesting thing is that the only two 32-bit architectures left supported are armel and armhf. Debian has already announced that this will be the last release that supports armel (https://www.debian.org/releases/trixie/release-notes/issues....), so I guess it'll be a matter of time before they drop 32-bit support altogether.
Yeah, mainframes are all about cases where the uptime is critical. Most modern systems are good with offering 99.9% or 99.99% reliability, with the understanding that trying to offer more than that just gets more and more expensive. Well... spending huge amounts of money on mainframes is one of the strategies to get to reliability numbers like 99.9999999%.
The legitimate usage is typically for workloads like relational databases (which are anyway single-machine architectures) that experience heavy load 24/7, part of fragile architectures that cannot tolerate downtime (remaining fragile as such because the original source code has been lost etc.), where "the system is down" causes tens of thousands if not hundreds of thousands of people to stop work.
Both Power and z are many billion dollar businesses each. Banking and other high finance is the stronghold for both. IBM still seems proud of z, Power seems merely tolerated these days which is a shame because it is a nice ISA and the systems are very nice too.
I have used Debian starting sometime around slink. I still type "apt-get ..." and it still works reasonably well. There have definitely been hiccups in upgrades over the last 25+ years but the amount of time/effort dealing with those is almost nothing in comparison to other linux and non-OSS systems I've dealt with over the same span of time. My only regret is not contributing more to the community.
The thing I like most about Debian is that you need to know at least a little about what is going on to use it. For me, it does a good job of following "as simple as possible and no simpler."
My first Debian install was in 1996. I had no real idea what I was doing, but it was amazing to me that I could remote-display windows from machines across campus, and it was alien compared to the windows 3.x/95 I was used to at that point. There was no apt at that point, or none that I was aware of, and adding new stuff was painful.
I started using debian preferentially as my workstation/desktop OS in about 2005, and was installing it on embedded systems (linksys nslu2) to make micro servers by … etch I think it was.
By 2008 I was at IBM and they allowed a choice of windows or redhat on your laptop, and if you were adventurous there was experimental support for Ubuntu which might work on Debian. I made it work and discovered that among 330k people there were 22 of us running it!
Always loved it, it always just made more sense than other distros somehow. My daily driver is a Mac now, but I still have a few Debian machines around.
As an owner of two i386 systems (both netbooks built around Intel's Atom N270), that run Debian, I am a little sad. I understand the reasoning, and I won't deny it is a very niche platform by now. But I had hoped Debian, with a history of supporting a wide range of platforms, would keep i386 going for a while longer.
Fortunately, bookworm will continue to receive updates for almost 3 years, so I am not in a hurry to look for a new OS for these relics. OpenBSD looks like the natural successor, but I am not sure if the wifi chips are supported. (And who knows how long these netbooks will continue to work, they were built in 2008 and 2009, so they've had a long life already.)
EDIT: Hooray, thanks to everyone who made this possible, is what I meant to say.
On the rest, I use mutt+msmtp+mbsync, slrn, sfeed, lynx/links, mocp, mupdf for PDF/CBZ/EPUB,
nsxiv for images, tut for Mastodon and Emacs just for Telegram (I installed tdlib from OpenBSD
packages and then I installed Telega from MELPA).
Overall it's a really fast machine. CWM+XTerm+Tmux it's my main environment. I have some SSH
connection open to somewhere else at the 3rd tag (virtual desktop), and the 2nd one for Dillo.
One sits in my bathroom so I can browse random Wikipedia articles while I'm, uh, busy. The other one sits on my nightstand and plays audiobooks/podcasts when I'm going to sleep.
So nothing critical. But something they are still good at, and being very small makes them a natural fit for these use cases.
I can't speak for the other poster, but I like the idea a lot. Having tools with specific purposes means I can avoid using my phone for everything. No matter what games I play to remove notifications/interruptions/etc. it's always a distraction and easy to be distracted from whatever I originally intended to use the phone for.
I do use the phone for audible, but I started both uses before I had a smart phone (I was very late to the game), and I am a creature of habit. Plus the netbook has a bigger display, more storage, and a real keyboard (again, creature of habit).
# example:
udevadm test-builtin net_setup_link /sys/class/net/eno4 2>/dev/null
ID_NET_LINK_FILE=/usr/lib/systemd/network/99-default.link
ID_NET_LINK_FILE_DROPINS=
ID_NET_NAME=eno4 <-- note the NIC name that will happen after reboot
Here's a one-liner, excluding a bond interface and lo. Gives a nice list of pre and post change.
for x in $(cat /etc/network/interfaces | grep auto | cut -d ' ' -f 2 | grep -Ev 'lo|bond0'); do echo -n $x:; udevadm test-builtin net_setup_link /sys/class/net/$x 2>/dev/null | grep NET_NAME| cut -d = -f 2; done
The doc's logic is that after you've upgraded to trixie, and before reboot, you're running enough of systemd to see what it will name interfaces after reboot.
So far I have not had an interface change due to upgrade, so I cannot say that the above does detect it.
enoX should always stay stable, as it's the BIOS (in some ACPI table) telling that this device/port has this ID.
ensX means the NIC in PCIe slot X, but in your PCIe tree you can have PCIe bridges, so technically you could have multiple NIC in the same slot (what the BIOS declare as a slot), so there was a lot of breaking NIC naming changes over the years in systemd to figure out the right heuristics that are safe, enabling/disabling slot naming if there is a PCIe bridge, but just in some cases.
Also for historical reasons the PCIe slot number was read indirectly leading to some conflicts in some cases (this was fixed in systemd 257)
Debian's signature feature (upgrade from stable to stable under 15 minutes) shines here too.
My first system migrated in less than 10 minutes, incl. package downloads and reboot. It's not a beast either. N100 mini PC connected to a ~50mbps network.
If your sources file references the release name (e.g.:bookworm), you change them to trixie, then “apt update && apt dist-upgrade”.
or,
If your sources file directly reference distro-suites (e.g.: stable), you just “apt update && apt dist-upgrade” since stable is now pointing to trixie.
In the first reboot, you run “apt autopurge” to remove packages which are not needed anymore.
Yes, esp. if you are using Debian at infra and in production. I upgraded three servers in the morning while sipping my tea. Nobody noticed anything.
For RedHat family, this is nigh impossible, requiring hours of planning, downtimes, etc., and even then, it's not guaranteed to work.
If you prefer a rolling release, and won't use it on a server, Debian Testing got you covered (unless you are in a security sensitive environment). My systems are well-isolated from outside, so desktop systems can run Testing without issues.
Servers and exposed systems always run stable with security updates installed automatically, though.
I've been on Debian forever and love it to bits. But in an act which can only be described as batshit insanity, they have chosen to patch Python's pip3 in a manner which breaks the --prefix option. On Debian
pip3 install <whatever> --prefix=/usr/local
will install into /usr/local/local, so one has to use the prefix /usr. The same command on, say, OpenSuSE will install into /usr and break your system. Barking mad.
Certainly a terrible UX, but the motivation is clear: they're trying to get PEP 668 protections for older versions.
Virtual environments work a lot better anyway, honestly. (With a properly crafted `pyvenv.cfg`, it should be possible to convince Python that your /usr/local is a virtual environment, but I can't be sure offhand if there are any serious negative consequences of that.)
Lots of debian love in this thread and it's great to see. If you're so inclined I encourage you to donate to Debian. We're all better off the more support goes to an ecosystem and operation like Debian.
I have been using Debian Trixie for a few months in testing now, I can attest that its a great, stable operating system. Definitely better than Ubuntu in terms of user experience.
The only complaint on a fresh install is that Cinnamon seems to use a ton of CPU when there's a little moving thingy anywhere on the screen (a browser tab that has a loading icon in the tab list is sufficient). This is most noticeable when you have a VM without graphics acceleration (don't ask why in the world my job requires that). Graphics without acceleration is always heavy, but this is an extra process doing whatever on top of the actual load
Then my private laptop has had a bunch of graphic issues after upgrading to 13 (it manifests differently in a lot of applications and it changes when you pick a different desktop theme, not even sure how to describe it). The new pipewire (pulseaudio replacement, idk why that needed replacing) does not work properly when the CPU is busy (so I currently play games without game sounds or music in the background). The latter then also sometimes (1 in 5 times maybe?) crashes when resuming from suspend, but instead of dying, spams systemd which diligently stores it all in a shitty binary file (that you can't selectively prune), runs completely out of disk space, and breaks various things on the rest of the system until you restart the pipewire process and purge any and all logs (remember, no selective pruning)... Tried various things I found in web searches and threw an LLM at it as well, but no dice. I assume these issues are from it not being a fresh install, so no blame/complaint here really, just annoying and I haven't had these issues when doing previous upgrades. Not yet sure how to resolve, perhaps I'll end up doing a completely new install and seeing what configs I can port until issues start showing up
Surely these things are not a Debian-specific issue, but I haven't noticed something like that with either 11 or 12
Edit: oh yeah, and the /tmp(fs) counter is at 1 so far. I wonder how many times I'll have run out of RAM by Debian 14, by forgetting I can't just dump temporary files into /tmp anymore without estimating the size correctly beforehand
I've been testing Cinnamon on my Ubuntu machine recently. I was close to deciding to switch from Gnome but I keep running into an issue whenever I have long running tasks in Blender.
In Gnome, Blender becomes unresponsive but everything else is still usuable. In Cinnamon, the entire system becomes unresponsive.
> "trixie" includes numerous updated software packages (over 63% of all packages from the previous release)
Wow, I'm amazed a third of packages haven't seen an update in, ehm checks
> After 2 years, 1 month, and 30 days of development, the Debian project is proud to present its new stable version
I'm a fan of old software myself, in the sense that I find it cool to see F-Droid having a (usually tiny) package that is over 10 years old but it does exactly what I want with no bugs and it works perfectly on Android 10. I wonder if those 30% more commonly fall in the "it's fine as it is" category or in the "no maintainers available" category
> Users running i386 systems should not upgrade to trixie. Instead, Debian recommends either reinstalling them as amd64, where possible, or retiring the hardware.
In the abstract I'm a big fan of supporting me old machines forever, but I have to ask out of curiosity - what hardware is practical to run these days and only has a 32-bit processor?
I have a few old PCs (towers) here which don't support amd64 mostly Pentium 4-based.
They all still have DVD reader drives and are nice for ripping CDs. Despite the fact that the drives are nearing 20 years of age (machines are from ~2005) they still perform better than most “new” external drives. Of course one could also move the drives to a newer machine but many of them use the IDE connector which is not commonly found on modern systems. Also, modern cases typically don't account for (multiple) 5.25" drives.
The other use case is to flash microcontrollers. When fiddeling around with electronics there is always a risk of a short circuit or other error that could in worst case kill the attached PC's mainboard. I feel much safer attaching my self-built electronics to an old machine than to my amd64 workstation.
Due to their age, I think the old machines may not live much longer -- I fear not even 10 more years, some of my old 32-bit laptops have already failed. Hence even for me it does not make sense to try keeping up the software support. Maybe I switch them to a BSD or other Linux distribution if they live long enough but for now the machines run OK with Debian Bookworm (newly oldstable), too.
I tried to repurpose an old laptop I had lying around as a "lie on the couch and surf the web or watch youtube" machine. It was one of the last 32-bit only cpus (pentium m), so I installed Debian bookworm (12) on it. Unfortunately it turned out it couldn't even play youtube videos at 144p without stuttering. So I E-wasted the machine.
I suppose as some kind of headless home server it could still have been useful. OTOH for something that runs 24/7 a RPi would use a fraction of the electricity and still be a lot more powerful.
So yes, beyond nostalgia and some embedded/industrial usecases, it's hard to see a use for a 32-bit only PC these days.
In my case I have a couple of first gen 32-bit Atom netbooks that I use regularly for the same things I've always used them for. The hardware still work just fine so I see no reason to replace them.
The difference between Debian and Ubuntu is decreasing with each release recently. I was pleasantly surprised that Debian recognized all hardware components in my laptop released one year ago
out of the box.
Hardware support is good and UI is great! It feels snappier than Ubuntu, may be due to lack of snap and fewer services and applications installed by default.
I love Debian and have a tremendous amount of respect for the people who work on the project. I no longer use Debian, but I think it's vitally important to have an anchor Linux distribution which isn't overly reliant on a for-profit entity and is truly community driven.
As this also lists lines of code, it sounds more like sources plus packages. Think space that a full mirror (src + generic + arch specific packages) would need.
Indeed, this is the amount of space that a Debian mirror would need to host all Trixie packages. So it's the compressed packages total size, not the space it would take to have all packages installed simultaneously (which also happens to be impossible, because of package conflicts/alternatives and Debian supporting 7+ different architectures).
The 403GB figure represents the total size of all source packages in the Debian archive, not the disk space required for a typical installation which is usually under 10GB for a desktop system.
Linux mint technically still does, perhaps because they’re lagging behind Debian. AntiX does but that has its own tradeoffs. With full potential respect for your situation, if you’re not able to obtain 64 bit hardware, there’s nothing that wrong with not upgrading your OS for now
Bookworm will get updates for another year if I have read the Debian Wiki correctly so perhaps there is no need for immediate action as Bookworm becomes oldstable?
In addition to Mint and AntiX there is also Slackware (I use xfce rather than KDE) but adding software outside the (large) base install is not an 'apt-get thing' process.
Salix Linux, based on Slackware, does have package repositories with a reasonable selection of applications. Salix is based on the stable Slackware 15.0 so has nicely aged packages.
Void Linux also has an xfce4 based live i686 iso you can look at and decide to install from.
I also use xfce, my computer is an Acer Aspire One, limited in ... well everything, I did upgrade the storage to an SSD, but it's still an Atom CPU, so whatever is the lightest is the best.
Looking forward to 13! Debian 12 in combination with Pipewire is my go to daily use professional audio workstation for 2 years. Coming from Windows, there are no more forced updates anymore preventing me from doing my job. This is a releave. It works so good! Linux for professional audio is really an option now! Most high-end converters are connected via MADI or ADAT anyway, so there is no driver problem existent. Drivers are a consumer grade discussion..
I've kind of been using Debian 13 for awhile now (I'm on Unstable) and for me what's impressive is how polished a default Debian installation is these days. With Gnome, you literally can run it as is, no config needed. It just works.
That being said, I like Flatpak, so I installed it (was super easy and Flathub provides instructions), and I added a few Gnome Shell extensions (a Dock so my wife can find apps when she occasionally uses my laptop).
Debian gives you a feeling of ownership of your computer in a way the corporate distros don't, but is still pretty user friendly (unlike Arch).
I'd definitely install Debian Stable on a grandparents' computer.
I used to like Debian when configuring ALSA/OSS, XFree86 and such was
a source of nightmares. Thus, debconf as a middle layer mechanism to
handle several distinct architectures, setups and hardware was
a neccesity. Ditto with Yast2 on SuSE. By 2004-2005... not much.
Even a bare Slackware with KDE and KDEi (and even XFCE) can do tons of
work by itself by just adding an user and accepting the default group belonging
array by pressing 'up' at the prompt.
Heck, even OpenBSD, minus the volume automount, which can be
handled in a breeze with toadd or tray-app in seconds; and if you
are smart you can figure DBUS/FDo mount points and integrate
then with XFCE/Plasma/Gnome without too much issues (hotplugd
can handle device umounting if you set up doas.conf accordinly).
The rest? MESA and X.org will handle most of the graphics stuff.
Video and audio drivers are autodetected on
almost every GNU and *BSD.
Printers are often wireless bound so any assistant with
look it up fast and attach it to CUPS.
Still, I can't handle DPKG/APT's slowness, even if there are libre distros
as Trisquel with it. If they rebased their distro as a simpler Parabola
LTS release with either Mate or LXDE setups,
the user experience would be almost the same, but installing
packages would happen at a much faster pace.
Maybe a niche concern, but SDL2 is still in Trixie. The sdl2-compat layer (translating SDL2 APIs to SDL3) is in testing, where SDL2 also exists side-by-side with it and is intended to be used to test and verify that SDL2 apps that use it are actually compatible.
Night-and-day decision-making process compared to Fedora and Arch, which both replaced SDL2 with sdl2-compat, broke a bunch of SDL2 apps because sdl2-compat isn't actually SDL2-compatible yet, and sent everyone to yell at the SDL team about it.
Even Arch kept SDL2 libs in the repo as a fallback. Fedora jumped straight from SDL2 in the repo to sdl2-compat or nothing. The only way to get consistent SDL2 builds on F42 is by building SDL2 from source.
Debian 13 trixie includes numerous updated software packages (over 63% of all packages from the previous release)
I’m not familiar with the metric definition they use, but I’d be worried if close to 100% of the packages they included in bookworm hadn’t been updated in the roughly 2 years between releases.
I use Debian for most of my servers, so I’m sure there is a valid explanation of that phrase.
I don't know why you think it would be different. Are you concerned about security updates? That's not part of the metric, as far as I can see.
And even if it was?
If you look at the number of packages in Debian, only a small portion have CVEs. There are nearly 30k package sources, and an output of 60k binary packages.
Yet we only get a few security updates weekly.
Another example? Both trixie and bookworm use the same firefox ESR (extended release) version. Both will get updated when firefox forces everyone to the next ESR.
Beyond that, some packages are docs. Some are 'glue' packages, eg scripts to manage Debian. These may not change between releases.
Lastly, Debian actually maintains an enormous number of upstream orphaned packages. In those cases, the version number is the same (sometimes), but with security updates slapped on if required.
From my perspective, outside of timely and quick security updates, I have zero desire for a lot of churn. Why would I? Churn means work. Churn means changed stability.
We get plenty of fun and churn from kernel, and driver related changes (X, Wayland, audio/nic, etc), and desktop apps. And of course from anything running forward, with scissors, like network connected joy.
It's not uncommon for small software packages to go years between updates - either because they're a simple utility that's feature-complete and rarely needs bug fixes, or because they're data files (e.g. packages of icons or fonts) which might not need to change at all.
Debian stable is just that - unchanging between major Debian versions. They do however push security updates when necessary, so you're not missing out on those
Any chance you know how they manage that? Surely not every package in the repos is supported for the entire 2 year cycle, so if a vuln comes out after a major refactor, it’s surely not easy to backport the patch.
They auto-import CVE feeds into the security tracker, file bugs for Debian maintainers to fix the issues, curate the tracking data, coordinate with upstreams and other distros to get fixes and so on. Some more on the team web page.
Not if you want to remain on new Plasma, you can expect Debian to lag several minor versions behind.
I've found it pretty easy though to use some KDE components built from source on top of the standard Debian packages. Build with kdesrc-build, then have those binaries linked to from your ~/bin and you're set. It might get difficult if you want to rebuild some key components like plasmashell itself but I've been using locally built versions of Kate and Konsole without issue.
Bookworm stayed on Plasma 5.27.5 when Plasma shipped bugfix releases up to 5.27.12. Debian may have cherry-picked a handful of patches from there, but that's a lot of bugfixes missed on a release that was already super old.
Even at this point, Plasma 6.4 has been out for almost two months and 6.3 will not get any more updates ever. While everyone else is upgrading, Debian is going to be stuck on an already unsupported version for another two years or however much.
Debian is great for what it is, but you better hope you don't run into issues with your desktop environment because they will not be addressed.
There are options if you're willing to put in the work and run the risk of breaking things.
You can always not install QT or KDE packages and compile your desktop from source. It's a major pain in the ass but I did it for years. A side benefit is you can participate in testing and interact with KDE developers directly.
Another option is to go all FrankenUNIX and add neon sources to your apt cache. I've done similar but I don't recommend it.
Or you can just run unstable. Lots of people do. I did for a long time, and as long as you're willing to fix the package system occasionally it's not a bad experience. Certainly better than the two previous options.
I'm too impatient to use a non-rolling release distro like Debian as my main OS, that, by my standards, already starts a new distro version with some outdated packages. I admire Debian though and it is my favorite server OS.
To be honest, it was a miserable experience to install it on your main computer without anything else available to look for help in case of problems. It was also hard to really try it due to lack of drivers for current (at that moment) ADSL modems.
But here I am a crapload of years later, still loving it :-)
Raspberry Pi OS is a derivative and not straight up debian. It's not a released yet. A beta exists and looks like this one will support an in-place update
Apparently Trixie needs a larger boot petition than prior versions and therefore I have to do new installs on three of my homelab machines - and probably my proxmox machine too. What a headache.
That sounds concerning, especially since Debian had hard-coded (!) a tiny /boot partition for encrypted disks for a long time. This already caused problems quite frequently (you have to manually delete them quite frequently, and which inhibits your ability to revert after a kernel regression - which, hm, I have noted to have been relatively common for Debian 12 "Bookworm" compared to usual ... hopefully Trixie is better, but if it makes the kernel-management problem harder that's a bad sign)
I have been tracking Trixie on my Resolve workstation for the past couple of months. The only hiccup was that the latest kernel did not support the ondemand governor, so I had to build a custom kernel to fix that.
Nice, more chaos to unstable again. Sincerely happy because of the release and new stable, but also selfishly happy because now unstable starts moving again.
I don't like GNOME's engineering or license, but as far as the UI and design are concerned I don't think anyone has topped it (minus the hot corner I guess).
I suppose 'power-users' can put up with 'small-button' GUIs but I do think a general purpose GUI needs to be easy to use for people who aren't used to controlling mice and keyboards all the time.
So what is the actual difference. These release notes are not very clear. They just give version bumps. How can people get excited when you give them nothing to get excited about?
I can't believe we've come to such a high number, and a particularly lucky one at that
Alas it's still not suitable as a daily driver for the average home user and probably never will be. It is unfortunate that Ubuntu has to reign supreme in that regard.
Don't feed the troll, etc... But I just had to bite on this bit:
> Alas it's still not suitable as a daily driver for the average home user and probably never will be. It is unfortunate that Ubuntu has to reign supreme in that regard.
It's true that Ubuntu used to be the OOB ready version of Debian, which "just worked", while base Debian took look of fiddling to even have wifi working.
These day though I find the opposite to be true: Ubuntu does lots of weird things I don't want, and I have to "fiddle" to disable all that. A base Debian install however (ISO with firmware bundled), just works.
For me, Ubuntu is officially off my list of distros I bother spending my time on.
Two kids in 4 to 16 range, and two adults in 30 to 46 age ranges have been using Debian on daily basis for almost a decade now. At least three of them are pretty "average home user". There has been forced use of windows (since school and employers wanted), but for home use Debian has always been better due to less maintenance needs and no distractions.
You can install debian and ubuntu with same DE and you'd be hard pressed to find a difference apart from the theme unless you are a power user who knows what snap is.
In fact, Ubuntu has never been an especially user friendly distro. At the beginning it was just a debian that was installed with debian's experimental installer before they decided to use it in stable. Nothing more, nothing less.
If you wanted to find a distro that was making efforts towards beginners looking for Gui config tools, you had to look at Suse and Mandrake (now Mandriva).
The only specific thing Ubuntu did for beginners is sending CDs for free at a time when not everybody had fast internet connections and would look for paper magazine to come with CD/DVD. And they have stopped doing that a loooooong time ago.
>The only specific thing Ubuntu did for beginners is sending CDs for free
Assuming you are not malicious I will kindly help with your bad memory, Ubuntu had always very good proprietary driver support, this made laptops actually work and helped beginners. I also remember they had a graphical installer compared to Debian and for sure this was beginners friendly. Maybe some other distro offered easy way to install and come with proprietary drivers setup but I can't remember a deb based distro doing that.
Anyway you were wrong, the CDs were not the only thing made Ubuntu appeal for beginners, there were Linux magazines with CDs each month and they were not super expensive , my first linux was a Kubuntu 6.10 from a magazine and I am still running Kubuntu today though i ran Debian, Sidux, Arch, Mandriva, SUSE in the past when I had time to try different distros, compile custom kernels etc.
>Proprietary driver installation was the sole reason of existence of Linux Mint which was a fork of ubuntu, so your memory is incorrect.
I think your memory is incorrect, you might be thinking of video codecs and maybe Flash not proprietary drivers, since Ubuntu already had support for easy install of drivers before Mint.
Or unless the capacity of you harddisk is limited, filling up with huge snap packages.
Or you need something that is broken by snap.
I helped a user after thunderbird after an upgrade could no longer open PDF-s in okular. It turned out that the thunderbird dpkg had been replaced with a snap and I spent quite some time trying getting it to work, filed bug reports, etc, before giving up and installing from Mozilla until I replace it all with Debian.
> Alas it's still not suitable as a daily driver for the average home user
I think that's fine for Debian. Maybe even a good thing.
Debian supplies a rock solid base for many general purpose tasks. Ubuntu and other distros are free to package that up in a user friendly way, but as a technical user I want to be able to go upstream and get a basic Linux system without extra stuff.
> Alas it's still not suitable as a daily driver for the average home user and probably never will be.
Why not?
My family members need little more than a web browser, media player, and office suite. Debian Stable is very suitable here; arguably more so than other distros, which tend to require maintenance more often.
The installation is slightly easier (but still hard because of USB install) and the website has a more appealing design. Except from that what is better in Ubuntu for the average casual user? Proprietary blobs are now included in the default installer since version 12.
Writing this from my Debian system, it's a great distro that has been excellent to me as a daily driver. I switched to Debian 6 after Ubuntu went way downhill and haven't had cause to regret it.
I like Debian's measured pragmatism with ideology, how it's a distro of free software by default but it also makes it easy to install non-free software or firmware blobs. I like Debian's package guidelines, I like dpkg, I like the Debian documentation even if Arch remains the best on that front. I like the stable/testing package streams, which make it easy to choose old but rock-stable vs just a bit old and almost as stable.
And one of the best parts is, I've never had a Debian system break without it being my fault in some way. Every case I've had of Debian being outright unbootable or having other serious problems, it's been due to me trying to add things from third-party repositories, or messing up the configuration or something else, but not a fault of the Debian system itself.
>I've never had a Debian system break without it being my fault in some way.
Debian is great but I can't say this is a shared experience. In particular, I've been bitten by Debian's heavy patching of kernel in Debian stable (specifically, backport regressions in the fast-moving DRM subsystem leading to hard-to-debug crashes), despite Debian releases technically having the "same" kernel for a duration of a release. In contrast, Ubuntu just uses newer kernels and -hwe avoids a lot of patch friction. So I still use Debian VMs but Ubuntu on bare metal. I haven't tried kernel from debian-backports repos though.
> Debian's heavy patching of kernel in Debian stable
Needs citation.
Debian stable uses upstream LTS kernels and I'm not aware of any heavy patching they do on top of that.
Upstream -stable trees are very relaxed in patches they accept and unfortunately they don't get serious testing before being released either (you can see there's a new release in every -stable tree like every week), so that's probably what you've been bit by.
LTS has had major breaking changes in various areas in recent times too, virtio was badly broken at one point this year, as was a commonly used netlink interface. Hat tip to the Arch kernel contributors who helped track this down and chase upstream, as we had mutually affected users. The debian and ubuntu bug trackers were a wasteland of silence and user contributions throughout the situation, and frustratingly continued to be so as AWS, GCP and others copied their kernel patch trees and blindly shipped the same problems to users and refused to respond to bugs and emails.
You're right stability comes from testing, not enough testing happens around Linux period, regardless of which branch is being discussed.
It's not easy testing kernels, but the bar is pretty low.
One of the unsung praises of Arch is that it's turned thousands of users into testers. Before someone says "that shouldn't be the user's responsibility" I'm going to say I'm not so sure. We're all in this together. I'd rather deal with a bug or two on my desktop at home if it means it gets fixed before appearing in a distro that gets used for servers at work and causes issues there where the consequences are much higher.
> One of the unsung praises of Arch is that it's turned thousands of users into testers.
You can do that well enough with Debian's "testing" and "unstable" release channels. Aside from the few months leading up to a new "stable" release, which usually isn't a big deal (and fixing regressions in "stable" should then be a higher priority anyway). Just don't install it on systems that you actually depend on to keep working. But running it on your desktop at home that you only use to play and experiment with is just fine.
I have a similar experience. My not-so-tech-savvy brother also has the same laptop setup I do (arch+XFCE). He knows to yay -Syyu and it's usually never a problem. The recent upgrade there was the vlc package split problem so I told him to hold on upgrading and that I'd come and do it. While I needed to sit and filter and install the optional dependencies myself for my upgrade, a week later it was already figured out (based on user feedback I assume) and the usual yay -Syyu installed just the right optional dependencies.
I don't consider myself particularly adept with linux. I've only been running it daily on the desktop for the last few years and, aside from mucking around with TWMs, I've not done much poking about with the internals.
Despite the reputations, I've had far fewer issues on Arch-based desktop distros than back when I was rolling Ubuntu and Debian.
That said, Debian on a server every time.
Yeah same. I think the release cycle actually doesn't matter at all. The reason for it is that the majority of breakage are caused by components/extensions of gnome and kde and non-DE-yet-complex software in distros with a lot of those present out of the box, like manjaro, breaking backwards compatibility every other week.
When people switch to arch they typically set things up from scratch, end up choosing simple tools and avoid most of the unstable stuff distros push onto you.
The virtio bug bit me. I have one hostnode (debian) with one nic that gets passed through to a virtualized opnsense. Storage is two consumer nvme in raid0 as system disk. I was expecting downtime with this setup, but kernel bug was not on my list.
Bear in mind, LTS and ELTS are not Debian maintained.
The wiki has more info on this.
The folks behind Debian LTS and Freexian ELTS are all Debian members/contributors, and the Debian LTS changes end up in the Debian archive, while the Freexian ELTS ones are publicly available, just in an external archive.
https://wiki.debian.org/LTS https://wiki.debian.org/LTS/Team https://wiki.debian.org/LTS/Funding https://wiki.debian.org/LTS/Extended
https://news.ycombinator.com/item?id=44853371
I think they mean the LTS kernels, not Debian's LTS.
Yep, I mean longterm trees from https://www.kernel.org/, to be clear.
Yes, seems so, thanks.
AFAICT, the patches are here: https://salsa.debian.org/kernel-team/linux/-/tree/debian/lat...
Whether that qualifies as "heavy" or not is of course a matter of opinion, but it's not nothing.
IMO, considering the size and scale of the kernel (millions of lines of code, variety of architectures supported, # of subsystems and ridiculous amount of device drivers ), these patches might as well be counted as nothing. I'd say they're basically shipping a pristine kernel :D
Classic Linux user response. Jeez…
These days all of my “Debian” bare metal systems are technically running Proxmox, which I think is a relatively happy medium as far as the base Debian system goes — the Proxmox kernel is basically the Ubuntu kernel, but otherwise it’s a pretty standard Debian system.
I’ve thought about (ab)using a Proxmox repository on an otherwise stock Debian system before just for the kernel…
Same at $work all physical servers run proxmox VE (by policy), 90% of VMs are debian (cloudinit genericcloud), the rest misc linux and various windows.
The upstream kernel already backports enough regressions on its own to its stable releases, Debian's kernel team does not help them too much with that.
Which GPU, display server and compositor stack are you using?
Integrated Intel GPU and no graphical system, just KMS VT (text console). That's what made it so frustrating - only displaying a console should not result in kernel panics under CPU load! Admittedly, the experience was anecdotal and years ago and I heard Debian is doing less of a RHEL-style "frankenkernel" now.
drm/i915 was a pretty miserable experience for me on one machine. The Intel drivers for that chipset around the 5.3 kernel era weren't good, I recall lots of bug reports at the time. Below is one of the several issues that I was affected by
https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/673
Intel's integrated GPU driver team, actually all driver teams, had a period of frequent screw-ups a while back (five years ago? Time flies). They also borked e1000e driver in the same period.
On the other hand, I had and still have many Debian installations, some with Intel integrated graphics. None of them created any problems for a very, very long time. To be honest, I don't remember even any of my Intel iGPU systems crashed.
...and I use Debian for almost two decades, and I have seen tons of GPU problems. I used to write my Xorg.conf files without using man, heh. :)
Maybe you can give Debian another chance.
Yeah, I ditched Ubuntu Server after too many upgrade headaches. I manage 75+ VPS instances for app hosting, and it's nerve-wracking doing maintenance updates knowing there's a chance one won't boot after. That's easily an extra 1-2 hours per VPS just to get it back. Switched to Debian back in the 8.x days in 2015 and it's been smooth sailing. Never had it break unless I was the one who messed it up.
Me too. All the server software (postgres, caddy, bun, etc) I'm using runs just fine on Debian, and I never have had updates break something on my Debian servers.
The only thing I can say against Debian is that it tends to start new server software immediately after install, before I have a chance to configure it properly. Defaults are sane for most packages, but, still, it scares me a little. In that I like the Red Hat approach of installing and leaving it off until I decide to turn it on.
It is a well-known issue with probably less well-known solutions, cf. <https://unix.stackexchange.com/questions/723675/debian-ubunt...>
I think this is the recommended way to avoid autostarting services on Debian.Good pointer. I remember learning it, and then forgetting it. Probably more than once.
Still should be the default behavior.
Just have sane firewall rules and you are good. E.g. if I install openssh-server and it auto starts, it doesn't make it out of my machine because my nftables does not allow inbound on port 22. It's just knowing the default behaviour and adjusting your practices for it.
That is a workaround for a ridiculous issue.
A sane firewall won't protect you from privilege escalation from a local attacker. While unlikely, this is one more breach that could be exploited.
Debian bundles AppArmor profiles for most services. This will prevent an attacker from accessing outside the perimeter drawn by the AppArmor profile.
This is the "you're holding it wrong" response to a clear design issue.
Aren't firewall rules part of the "configuration" the OP talked about?
No, because you can install and configure the firewall before you install package X. (without knowing anything about X, your firewall defaults can just prevent X from doing anything)
But you can't (easily) configure package X itself before you install it; and after you install it, it runs immediately so you only get to configure it after the first run.
> And one of the best parts is, I've never had a Debian system break without it being my fault in some way. Every case I've had of Debian being outright unbootable or having other serious problems, it's been due to me trying to add things from third-party repositories, or messing up the configuration or something else, but not a fault of the Debian system itself.
You're not trying hard enough ;-)
I have Debian on an old MacBook Pro and had it on an even older iMac, and I've had a few problems over the years. Always with proprietary drivers - WiFi, graphics, webcams, etc. - Apple really don't want people using free software on their hardware. There's always been a fix, but there have been a few stressful moments and hoops to jump through.
But it's definitely my favorite distro, and I run it everywhere I can. Pretty much always "just works" anywhere but Apple.
I’m not trying hard enough. Feel the same as you and GP for two decades and counting.
> after Ubuntu went way downhill and haven't had cause to regret it.
In what way Ubuntu went downhill?
For me, it was a combination of Ubuntu breaking upstream and introducing its own unnecessary systems.
I had a few issues caused by Ubuntu that weren't upstream. One was Tracker somehow eating up lots of CPU power and slowing the system down. Another was with input methods, I need to type in a pretty rare language and that was just broken on Ubuntu one day. Not upstream.
The bigger problem was Ubuntu adding stuff before it was ready. The Unity desktop, which is now fine, was initially missing lots of basic features and wasn't a good experience. Then there was the short-lived but pretty disastrous attempt to replace Xorg with Mir.
My non-tech parents are still on Ubuntu, have been for some twenty years, and it's mostly fine there. I wouldn't recommend it if you know your way around a Linux system but for non-tech, Ubuntu works well. Still, just a few months ago I was astonished by another Ubuntu change. My mom's most important program is Thunderbird, with her long-running email archive. The Thunderbird profile has effortlessly moved across several PCs as it's just a copy of the folder. Suddenly, Ubuntu migrated to the snap version of Thunderbird, so after a software update she found herself with a new version and an empty profile. Because of course the new profile is somewhere under ~/snap and the update didn't in any way try to link to the old profile.
Then there were stupid things like Amazon search results in the Unity dash search when looking for your files or programs. Nah. Ubuntu isn't terrible by any means but for a number of years now, I'd recommend Linux Mint as the friendly Debian derivative.
Snaps? Proprietary package managers are never great.
As I understand it, snap the package format is not proprietary. Its as open source as say flatpak. What is proprietary is Canonical official snap store, and they patch their version of snap to only use that store. It'd be the same as flatpak being tied to only flathub.
Of course that goes against the spirit of FOSS, but there's a bit more nuance there than simply saying "snaps are proprietary".
Snaps don't just suck from an ideological but also practical perspective, as described for Thunderbird. Firefox on Ubuntu has also serious permission issues with webcam support OOTB even experts are struggling with (involving AppArmor, pipewire, snap, and FF device config). and has become unusable for things like browser-only MS Teams on mainstream notebooks.
Containers, popular as they may be on servers, can only add breakage and overhead to desktops, especially for an established and already much better organized system like Debian's apt. There just haven't been any new desktop apps for way over a decade that would warrant yet another level of indirection.
In addition, applications under snap are much slower to start. That's just not acceptable.
I've tried making snap packages, but I discovered they're very tightly tied to Ubuntu's base packages. They're not portable at all. In essence they're effectively just a secondary Ubuntu-specific package format for user-level applications.
For example, with flatpak you select a base runtime for your package that contains mostly system-agnostic libraries. With snap, you specify an Ubuntu version as a base runtime and additional dependencies that are Ubuntu packages.
My understanding is that the base layer (similar to what FlatPak provides) is shared and downloaded by the snap manager so it is portable as long as you want to download it.
The end result should be similar to FlatPak where you have practically no dependencies as it should package almost everything.
Did they release the server components for hosting your own snap repositories, yet?
I can't seem to find it. Any pointers would be helpful, so at least I can know the latest state of this thing.
Snapd still hardcodes Canonical's snap store signing key and provides no mechanism to add your own keys. Any other snap repos will be treated as second class citizens.
No, but it's trivial to implement since Snap is open source so you know exactly what sort of payload it wants.
> If I try to manipulate what you are talking about, I can attempt to frame something as open source which isn't.
I didn't say "the snap format".
The server isn't, and the client is hostile to using an alternative server. Snaps are a solution, and picking out one piece is deceptive.
I don't really understand why this is such a big problem. You don't have to use snaps.
Defaults matter a lot, and snaps are the default in Ubuntu.
The topic is not whether snaps are avoidable or not, but the Ubuntu is going downhill. And snaps are purported to be part of that downhill, which would be Ubuntu's NIH syndrome. As far as I know, Ubuntu's only successful development is Ubuntu itself - the other projects have all failed over the years, and snap, while ongoing, is not winning any popularity contests either.
Snaps per se are no better or worse than flatpak. Canonical's mistake, IMO, was to make their store the only place snaps can be hosted. That is the "proprietary" bit everyone keeps talking about.
But in practice even for flatpak the only realistic place you can publish your flatpak if you want any traction at all would be flathub, so both formats have only one store right now. But flatpak allows a custom store while for some strange reason Canonical decided not to allow snap that freedom.
Another problem is, Canonical promised to release server components and enable alternative stores, and just forgot that they made that pledge.
Also, rugpulling users and migrating things to snaps without asking their users in order to "create a positive pressure on snap team to keep their quality high" didn't sit well with the users.
> But in practice even for flatpak the only realistic place you can publish your flatpak if you want any traction at all would be flathub
But, for any size of fleet from homelab to an enterprise client farm, I can host my local flathub and install my personal special-purpose flatpaks without paying anyone and thinking whether my packages will be there next morning.
Freedom matters, esp. it that's the norm in that ecosystem.
I was neutral-ish about Ubuntu, but I flat out avoid them now, and migrate any remaining Ubuntu server to Debian in shortest way possible.
I'm using Debian for the last 20 years or so, BTW.
Yes, same. I started with Ubuntu back in the day, because the server I inherited ran Ubuntu, and it was just natural after that for me to run it on the desktop as well. I grew to dislike their NIH over the years, tried distro hopping, and settled on Debian.
Yes, I agree. Snaps or Flatpak, not much of a practical, technological difference. What sets them apart is the way the distribution is handled, including the open source availability of the backend, which enabled for example Red Hat and Elementary to run their own stores.
If you are making your own distro, creating your own flatpak store is trivial, that's all what matters. Linux Mint doesn't use snap exactly because Canonical forces everyone to use their snap store.
Canonical doesn't force anyone to use anything. Snap is open source, just modify it to use a different store if you want. Mint literally forked a zombie DE, but changing a few lines of code in snap is an issue...
Defaults matter a lot, snap is not open source (client is, backend isn't), you cannot "just modify it (Ubuntu)" to use a different store, because Ubuntu installs snaps even with apt. Mint is not part of the discussion.
> Mint is not part of the discussion.
Read the parent comment I responded to
Mea culpa, I glossed over that!
https://news.ycombinator.com/item?id=44849691
> which would be Ubuntu's NIH syndrome
Red Hat do the same. They reinvented the wheel on multiple occasions (systemd and it's whole ecosystem like systemd-resolved and timed and the whole kitchen sink; podman, buildah, dnf, etc etc.)
They just have more success on getting their NIH babies accepted as the standard by everyone else. Canonical just fail at that (often for good reasons, Unity was downright crap for some time) and abandon stuff, which doesn't help their future causes.
Canonical did their own NIH init daemon called Upstart which failed due to the fundamental design and the implementation being plain bad. Redhat builds better software which is why their NIH gets more adoption.
ChromeOS still uses upstart.
Upstart came before systemd; much of the reason for systemd’s creation was fixing what was considered fundamental design mistakes in Upstart: <http://0pointer.net/blog/projects/systemd#:~:text=On%20Upsta...> (under the heading “On Upstart”).
> systemd
https://bbs.archlinux.org/viewtopic.php?pid=1149530#p1149530
> like systemd-resolved and timed
They're not forced on anybody, they're not required by systemd, and many distributions use more feature-rich alternatives (including, afaik, RHEL — last time I looked at it, they used dnsmasq and chrony). They're also often shipped as separate optional packages:
> podman, buildahStill not anywhere near as popular as Docker. Although technically they're far better than Docker, and if anyone is using them, it's for that reason.
> dnf
Only used by RHEL and its upstream Fedora?
---
All of this makes very little sense.
>> podman, buildah
> Still not anywhere near as popular as Docker. Although technically they're far better than Docker, and if anyone is using them, it's for that reason.
NIH packages are generally expected to be less popular, yes. They have some technical merit, though in my opinion that's mostly trade-offs rather than one being strictly better than the other. I would be surprised if everybody using them is using them because of technical merit as opposed to it being pushed by the distro.
Red Hat builds really good stuff. NIH is sometimes right because nobody invented the stuff at all. Standard Unix tools are great but they don't solve everything, so we've ended up with most distros having "the Debian way" or "the Red Hat way", the main difference of course being deb/apt/dpkg vs rpm/yum/dnf. When building an embedded system with Yocto, the basic choices are also Debian or Red Hat style, though you can of course do anything.
Special mention goes to NetworkManager, which has become the de facto standard way to configure networking because it's good. And with nmcli I can even remember how to connect to wifi from single user mode.
>They just have more success on getting their NIH babies accepted as the standard by everyone else.
This depends on the phrasing. We could also say that Red Hat produces actually useful software, in contrast with Canonical, whose developments don't seem to provide value over existing solutions.
We could also say that Canonical tries really hard to do exactly what Red Hat does, but in a slightly different space, and not very successfully.
A major difference is that Canonical projects have copyright assignment policies, while Red Hat projects don't - this probably explains a lot of the difference in adoption dynamics.
You're right. You don't have to use snaps. Ubuntu migrates packages slowly in behalf of you.
Using apt to install some packages installs snap plumbing and downloads the package as a snap automatically. You don't have to install it manually.
There's no malicious intent though, it's made to "impose a positive pressure on the snap team to produce better work and keep their quality high" (paraphrased, but this was the official answer).
Installing the inferior snap packages when you apt get is one of the worst cases of a Linux distro refusing to respect the user's intent that I've experienced.
Preach.
This has got to be the most user-blind self-imposed preference in a modern operating system outside of Microsoft's BS.
If you're going to use an OSS operating system, the control of what is placed on the system should be inherently with the user. If the developer has a question if a new package should be added or is required, throw a prompt and ask -- with a default to not use application containers and the default packaging system.
Really not hard.
And one of these migrations broke my workflow substantially enough that a dist-upgrade turned into a complete system reformat to Debian and cost hours that I couldn’t afford.
Debian has been a safe haven since.
You really have to work to avoid them; ex. `apt install firefox` will install the snap
You sort of do. It's really hard to avoid them, because they've modified "apt" to install snaps by default without asking.
I'm with my neighbor comments. How do you use Ubuntu without snaps? The base Ubuntu install already comes with several snaps. Installing random things through apt leads to snaps. I personally do not know how to avoid snaps on Ubuntu.
I made this snap "alternative" to solve this exact problem: https://github.com/justinclift
Packaged as: https://github.com/justinclift/snapd-empty/releases/download...
It's just an empty package that tells the system snap is installed, to stop the broken dependency chains you otherwise get from force uninstalling snap.
It's been working fine on a handful of Ubuntu 24.04 systems I've been handed and can't change the OS of, for about half a year now.
Ugh, I somehow managed to paste an incorrect repo url in my comment above.
It should be this: https://github.com/justinclift/snapd-empty/
You migrate to Debian. Everything else is a bandaid that can be rug pulled any time Canonical feels like doing so.
I've done that some years ago and couldn't be happier.
You uninstall all snaps, uninstall snapd, and block it from being installed via APT.
Then you add e.g. the mozilla PPA such that its firefox package gets installed instead.
Yes you do. Some packages aren't available anymore in apt
If you use Ubuntu, yes you do. It’s why I ditched Ubuntu
I forget the exact context but I recall an Ubuntu dev stating they have more users of the Firefox snap alone than trendy distros have entire users.
I think it’s worth keeping that in mind with all the hate Ubuntu gets. Most users are just silently getting their work done on an LTS they update every two years.
Well, I don't know which trendy distro the dev is referring, but Debian is complete opposite of Trendy. It's a bedrock distro silently running almost everywhere in some form or shape.
Most of the Linux-based (enterprise and/or embedded) appliances are built upon Debian, for example.
P.S.: The total number of Debian installations and their derivatives are unknown BTW. Debian installations and infra do not collect such information. You can install "popularity-contest", but the question defaults to "no" during installation, so most people do not send in package selection lists, unlike Canonical's tracking of snap installations.
They have also had more malware on the snap store than all other distros have had in their official repos combined.
IMHO, it also became too complex, with too many things installed by default and too much upstream patching.
This made it more fragile. It was really nice in the late 2000s, but gradually became worse.
all the weird proprietary Canonical stuff they try to put into vanilla Debian and have it replace common stuff.
snap, lxd (not lxc!), mir, upstart, ufw.
It's neverending, and it's always failing.
LXD was forked as Incus, and it’s an absolute delight.
Seamless LXC and virtual machine management with clustering, a clean API, YAML templates and a built-in load balancer, it's like Kubernetes for stateful workloads.
Incus is fantastic. I think Proxmox is where everyone is migrating to after the VMWare/Broadcom fiasco, but people should seriously consider Incus as well.
Snaps, and ads in the motd
Plus reduced support duration to increase adoption of Ubuntu Pro. Changing some packages ever slightly so they behave a little differently.
Switch to sudo-rs, uu-coreutils (rust based stuff), etc., etc.
It's not a Debian derivative anymore. It's something else.
Was not my cup of tea before, it's even more not my cup of tea now.
Switching to rust-based system software is very different from the clearly profit seeking (or control seeking which is just long term profit seeking) changes like ads and snap (with massive friction to not using snap).
Yes, but I prefer glibc + GNU Coreutils based systems in my installations. They're additional nails on top of the (fatal) ones like snap, Ubuntu Pro and MOTD ads.
The alternative question to ask is: in what way has it gone uphill versus just using Debian?
In the early days it had a differing and usually better aligned release schedule for the critical graphics stack.
As a function of time, you are increasingly likely to get rug pulled once Shuttleworth decides to collect his next ransom.
> in what way has it gone uphill versus just using Debian?
Their lawyers' willingness to risk shipping pre-built zfs kernel modules (that are always in sync with the kernel). Pretty important if you're into that sort of thing, it's easier to remove cruft once post-install than to keep an eye on DKMS for years (making sure that it hasn't disassembled itself and continues working).
Minutes to start Firefox on one of my machines.
Amazon ads in the Unity application menu (what was it called, 'lenses' or something?).
I'm an old-school user so I'm not exactly Ubuntu's target audience, but for Ubuntu was bad about a lot of the older, lesser-used bits of Linux.
The two things I can remember were problems with NFS out of the box (outside having to install nfs-common, which I'm fine with) and apt-cache not displaying descriptions of packages. There were lots of other, minor annoyances that affected people like me but wouldn't affect someone who got into Linux desktops after, say, 2010. My memory sucks though so those are the two I remember. Yes, there were bug reports filed and yes, they sat in the tracker for years with no attention from Ubuntu.
I wound up back on Debian once I got old enough that I didn't care about being behind the times a couple years.
Oh....snap.
> I like dpkg, I like the Debian documentation even if Arch remains the best on that front.
That's curious, because when I was learning to make Debian packages, I found the official documentation to be far better than I had seen from any other distro. The Policy Manual in particular is very detailed, continually improving, and even documents incremental changes from each version to the next. (That last bit makes it easy for package maintainers to keep up with current best practices.)
Does Arch have something better in this department?
Are you perhaps comparing the Arch wiki to Debian's wiki? On that front I would agree with you.
> I've never had a Debian system break without it being my fault in some way.
My experience has been contrary to that. I'm a Linux user of 25+ years with various distros but about half of that time with Debian as my main desktop. I broke up with Debian about ten years ago thinking we could still be friends, but every time I've tried to put it on a new box it since then something weird has happened, most recently about a month ago on a completely new Intel N150, when it gave me some stick about video modes. Today my laptop got hosed by an attempted upgrade from bookworm to trixie, as in tons of error messages and then no more docker and no more virtualbox. No harm done because Debian taught me long ago to store a copy of the whole root filesystem on external media before an upgrade, but now the clock is ticking until I have to migrate off it or get stuck with something too old to be compatible with anything.
> I like Debian's measured pragmatism with ideology
There is plenty that could be said of Debian but as far as I’m concerned that’s not part of it.
Debian patches software for purely ideological reasons because they think they are not free enough. That’s not pragmatism. That’s the reverse of pragmatism. It certainly is a real drag on the teams developing the software they try to ship.
You weren't around for when they broke the OpenSSL random number generator for no good reason. That was back in 2008 and it created vulnerabilities that persist to this day. https://16years.secvuln.info/
I still use Debian but it's hard to forget stuff like that even after all these years.
Though I agree with you, if that's the bar you're setting then Debian comes out far ahead of any other OS that I've ever used - Linux based or not. I can recall dozens of worse Windows bugs, most of which did not even affect me because I was not using Windows at the time. Mac has its share too.
What do you expect Debian to do today about this 17 years old incident?
> And one of the best parts is, I've never had a Debian system break without it being my fault in some way.
https://blog.kronis.dev/blog/debian-updates-are-broken
https://blog.kronis.dev/blog/debian-and-grub-are-broken
Then again, I’ve had most software occasionally break, I’m thankful that Debian exists.
Debian is my foundation. I keep servers on Old Stable and test new release features on an ephemeral system.
I learned nftables with Bookworm and labwc with Trixie.
labwc supports Wayland with Openbox configuration.
Why do you remain on old stable instead of stable?
I think this is all true, but the "being my fault" part has gotten better for me with nixos. Broke it? just reboot into the previous version and get configuration.nix back from git. I had to reinstall exactly once in 2016 shortly after the first install, but I don't know what I did wrong. the third time I installed nixos was last week when I bought a new computer that came with Windows.
I've been always a Fedora person, still am. But my PC moved to Proxmox (debian) in 2023. Now a Fedora Atomic sits in a VM running flatpaks and podman containers :D
You don't mention say what you like specifically about Debian, most of what you wrote could be said for a lot of distributions.
So here is what I _don't_ like about Debian :-)
- I don't like Debian package tooling (dpkg, debootstrap, de build...). Actually I hate everything about the experience of Debian packaging. Every time I package for Debian, I end up with a messed up setup of chroots and have to make triple sure nothing leaked from my environment.
- Debian has a habit of repackaging everything at their own sauce, disregarding upstream philosophy. Debian packages will have their own microcosm of configuration directories, defaults, paths, etc. orthogonal to what a pristine installation look like.
- Debian has the annoying habit of default starting installed services. So you always have to dance around your configuration management to disable services, install them, configure them, then restart them.
Do you usually update in place or do a fresh install whenever a new major version comes out?
I always update in place. And I follow all the upgrade procedure advice in the release notes.
Thank you to all the Debian volunteers that make Debian and all its derivatives possible. It's remarkable how many people and businesses have been enabled by your work. Thank you!
On a personal note, Trixie is very exciting for me because my side project, ntfy [1], was packaged [2] and is now included in Trixie. I only learned about the fact that it was included very late in cycle when the package maintainer asked for license clarifications. As a result the Debian-ized version of ntfy doesn't contain a web app (which is a reaaal bummer), and has a few things "patched out" (which is fine). I approached the maintainer and just recently added build tags [3] to make it easier to remove Stripe, Firebase and WebPush, so that the next Debian-ized version will not have to contain (so many) awkward patches.
As an "upstream maintainer", I must say it isn't obvious at all why the web app wasn't included. It was clearly removed on purpose [4], but I don't really know what to do to get it into the next Debian release. Doing an "apt install ntfy" is going to be quite disappointing for most if the web app doesn't work. Any help or guidance is very welcome!
[1] https://github.com/binwiederhier/ntfy
[2] https://tracker.debian.org/pkg/ntfy
[3] https://github.com/binwiederhier/ntfy/pull/1420
[4] https://salsa.debian.org/ahmadkhalifa/ntfy/-/blob/debian/lat...
The maintainer has a short explanation here: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1098866#10
> The webapp is a nodejs app that requires packages that are not currently in debian.
Since vendoring dependencies inside packages is frowned upon in Debian, the maintainer would have needed to add those packages themselves and maintain them. My guess is that they didn't want to take on that effort.
> but several features in ntfy won't be available through debian packaging due to missing golang and nodejs packages
Woah. Shouldn’t Node and Golang be in Debian’s official repos by now?
Nodejs itself is, but when you install a node project manually, you type npm install and wait while it downloads the 500 different packages it depends on.
Debian follows the same philosophy as for other more traditional languages and expects that all these dependencies are packaged as individual Debian packages.
Just jumping in to say that this is making me genuinely reconsider adopting a licence/policy that forbids repackaging: the fact that someone can repackage my project, but worse, and still use my project's name? Absolutely not. I do not want the burden that inevitably comes when people complain to me that this or that is missing from a repackage.
I mean, that's just how OSS works. Anyone can fork your thing, do whatever, and call it a day. Going MIT or whatever won't save you either - this repackaging business is basically the entire business model of AWS.
Forking is fine, do whatever, but as soon as you make actual changes to the code then adopt your own name. This is what trademarks are for, it's just that official trademark registration is somewhat inaccessible (eg: cost) for open source projects. Could you imagine trademarking every little project you make just in-case it gets repackaged by someone who tears huge chunks of it out?
Yes but not all packages written in those languages are.
On the web part:
Debian sources need to be sufficient to build. So for npm projects, you usually have a debian-specific package.json where each npm dependency (transitively, including devDependencies needed for build) needs to either be replaced with its equivalent debian package (which may also need to be ported), vendored (usually less ideal, especially for third-party code), or removed. Oh, and enjoy aligning versions for all of that. That's doable but non-trivial work with such a sizable lockfile. If I would guess the maintainer couldn't justify the extra effort and taking on combing through all those packages.
I also think in either case the Debian way would probably be to split it out as a complementary ntfy-web package.
Just wanted to say thanks for ntfy! I use it daily to notify me on events from my home Meshtastic node.
Thank you for ntfy, it's such a useful piece of software!
It might be a better idea to release this as a container (if it isn't already) to take care of the dependencies.
https://docs.ntfy.sh/install/#docker
> The ntfy image is available for amd64, armv6, armv7 and arm64. It should be pretty straight forward to use.
> As a result the Debian-ized version of ntfy doesn't contain a web app (which is a reaaal bummer), and has a few things "patched out" (which is fine).
My advise to you is to deny all support from people using the Debian version of your software and automatically close all bug tickets from Debian saying you don’t support externally patched software.
You would be far from the first to do so and it’s a completely rational and sane decision. You don’t have to engage with the insanity that Debian own policies force on its maintainers and users.
I agree that maintainers should not be expected to support patched versions of their software, but as a user I like the Debian policies you call insane. I would actually pick Debian exactly because they are cautious with the dependencies.
Debian is not cautious with the dependencies. Debian breaks a lot of what they ship, sometimes flagrantly like removing a whole feature, sometimes insiduously by introducing new bugs. I don't really care that Debian doesn't view it as breaking things. From my point of view, users trying to get my product get subpar experience in a way which is far from explicit.
I personally wouldn't use Debian but people are free to do whatever they want. I don't want to waste my time dealing with Debian maintainers and how they think software should work however. I advise all software developers to do the same and am vocal about it because it's easy to get guilt tripped in the idea that you should somehow support their users because they want to use your product or that introducing changes to support their esoteric targets somehow make sense because they have done the work despite the burden of futur support actually landing on you.
I want to make clear to people who decide they have no interest in it that they are not alone and it's perfectly fine.
And to be clear, I am singling Debian here because they are by far the worst offender when it comes to patching but the comment applies equaly to any distributions that apply invasive patches.
Debian IS more cautious with dependencies, in that you won't get hidden dependencies that aren't in the repos.
I don't want to install an app that downloads and executes 500 node packages that I don't know what they do. Those packages should already be vetted and in Debian. If not, then I'm not interested.
Side stepping the distro repos for dependencies for software in the repos leads to unexpected behavior.
> Debian IS more cautious with dependencies, in that you won't get hidden dependencies that aren't in the repos.
For a definition of cautious I don't personally share.
Debian doesn't vet packages. Debian maintainers are less competent than the "upstream" they question approximately all the time, which is why they keep breaking stuff in more or less severe way (OpenSLL anyone?). And let's not even talk about the insane stuff like when maitainers decide to support a fork they like instead of the piece of software users actually want (Libav anyone?).
> If not, then I'm not interested.
And that's your choice. That doesn't mean developers should care, nor that it is actually a good idea.
ntfy is a very useful tool. Thank you very much for making it and also for maintaining the ntfy.sh service for those of us too lazy to self host.
Congratulations!
Debian has been the stable footing of my Free computing life for three decades. Everything about their approach — from showing me Condorcet, organising stable chaos, moving forward by measured consensus, and basing everything on hard wrought principles — has had an effect on me in some way, from technical to social and back again.
I love this project and the immeasurable impact it has had on the world through their releases and culture.
With all my love, g’o xx
> i386 is no longer supported as a regular architecture: there is no official kernel and no Debian installer for i386 systems. The i386 architecture is now only intended to be used on a 64-bit (amd64) CPU. Users running i386 systems should not upgrade to trixie. Instead, Debian recommends either reinstalling them as amd64, where possible, or retiring the hardware.
Impressive that i386 support made it all the way to August 2025. I have Debian 10 Buster running on a Pentium 3 which only EOL'd last year in June 2024. It's still useful on that hardware and I'm grateful support continued as long as it did!
OpenBSD still supports i386 for those looking for a modern OS on old 32-bit hardware.
Hopefully i386 (or perhaps a new i386-like port with added support for 64-bit time values) can move to the unofficial Debian Ports infrastructure for Debian 14 (forky) or Debian 15 (duke). Debian Ports has a m68k port, so supporting one for i386 shouldn't be a huge problem.
To what end? Outside of sheer nostalgia if you are running ancient hardware, you probably have a bespoke application which requires that environment. Either you cannot change for hard technical, compliance, or just fear of the unknown. Firewall it from the internet and continue to run whatever release last worked.
I am not happy about unnecessary ewaste, but an i386 almost certainly has and order of magnitude less horsepower than a raspberry pi or N100.
Debian's tagline is the "universal operating system". It's a distribution with active ports on a very large number of architectures [1], even incredibly obscure ones.
The goal of universal compatibility that separates the Debian project from commercial software and even other open-source projects.
The legacy x86 architecture is still far more popular than some that platforms that Debian advertises as having official support for and there has been x86 based processors manufactured for niche applications until recently, eg, AMD Geode and others.
I find it really unfortunate Debian Project is removing official support for new x86 installations. The silver lining is it seems like they'll be an unofficial port and it's likely niche distributions like MX Linux and AntiX will maintain their own builds.
It would be ideal if open-source can develop stronger mechanims to keep support for the large numbers of these relatively niche architectures (eg, through increased usage of emulation over real hardware).
[1] https://wiki.debian.org/SupportedArchitectures
My Linux machine is very modern, but I still need i386 architecture support installed, because Steam requires 32-bit support. And Steam requires 32-bit support so people can play 15-year-old games.
(Admittedly, the 32-bit support Ubuntu ships is less than a full OS and you can't install Ubuntu on a 32-bit machine these days)
Debian is doing basically the same thing Ubuntu is with regards to i386. Packages are still being built for the architecture, but i386 systems aren't supported, and there's no 32-bit kernel package.
e.g. notice that i386 is still listed at the bottom of https://packages.debian.org/trixie/bash
So you have an amd64 CPU and Debian's "i386" packages will keep working on it. As per the release notes:
> The i386 architecture is now only intended to be used on a 64-bit (amd64) CPU.
Maybe it's one of those games that runs too fast if the CPU isn't clocked at 33 MHz. ;)
It would probably take a few days to start Steam on one of those considering its load times on current hardware.
I don't think this is the case anymore, see https://gitlab.winehq.org/wine/wine/-/releases/wine-9.0#wow6... though it's not the default everywhere.
According to Passmark the Pentinum 4 1.3Ghz is 55 times slower than a Raspberry Pi 5, so I'd guess it's at least two orders of magnitude. The original Pi is 16 times faster than a P4 1.3Ghz...
You can recycle e-waste (and yes, I know SOME e-waste ends up in China/India/etc. Not all does.)
The e-waste is of substantially less concern than the massive difference in carbon footprint from power consumption.
> The original Pi is 16 times faster than a P4 1.3Ghz...
It isn't. Had both.
~700MHz mostly in-order ARM 11 without SIMD. Worked with these a lot.
..vs..
1.3GHz out-of-order with SIMD
The i386 architecture hasn't been dropped, it is still available in the archives to support 32-bit applications. The major change is that there no longer is a 32-bit kernel in the archive (the package linux-image-686 is no more). But most packages are still available in their i386 versions:
It still exists but without any official iso or installer.
If that's all there's to it, you can still use debootstrap, compile a kernel, and point the root parameter to your shiny new install.
If the official i386 arch was built with instructions that your hardware doesn't support, tough cookies.
If the official i386 arch was built with instructions that your hardware doesn't support, tough cookies
While theoretically possible, that would only happen on processors older than 30 years. Debian's i386 architecture still uses -march=i686 as its baseline compiler target, which is the venerable Pentium Pro: https://en.wikipedia.org/wiki/P6_(microarchitecture)
I have AMD Geode hardware circa 2007 (18 years old) that only has partial support for i686. Requires a true 3/4/586 kernel.
Not sure why you are downvoted, I guess people don't believe this is true. To confirm: The AMD Geode LX was a <5W 32-bit x86 processor which did not support SSE instructions, and is therefore not fully i686 compatible. According to Wikipedia, it was produced until 2019: https://en.wikipedia.org/wiki/Amd_geode#AMD_Geode
It was used in the OLPC XO-1. The Cisco ASA line of firewalls also used Geode processors at least at some point in its lifetime.
In case anyone wants to do that, here is the doc for new ports:
https://wiki.debian.org/PortsDocs/New
Much of that information has to do with creating a new hardware port from scratch. The i386 support just needs to be "demoted" to the Debian ports infrastructure once it's officially scheduled to get dropped from the main Debian repository (which could well happen starting either in Debian forky or duke), and this can probably be done with some special handling.
(Answering the "to what end?" question, a lot of 32bit-only hardware is still available and dirt cheap in the second-hand market (e.g. early "netbooks"), much of it quite well-built and enjoyable to use. While such hardware can no longer realistically browse the "modern" web, it can still find a lot of use for more lightweight tasks, including acting as a "thin client" for more powerful machines.)
Well, the existing i386 port is going to remain as-is for supporting old software (especially games) (but the CPU baseline will likely get increased), it isn't going to be removed, so you would need a new architecture for 32bit-only hardware.
Since i386 is not going to do the 2038 transition either (since that would break the ABI), also you would need to either make a new ABI for the new port, or do the 2038 transition for it too.
Over time more and more 32-bit bugs will get introduced, so there will be lots of maintenance work to do too.
> ...it isn't going to be removed
I had seen conflicting information about this, though nothing official. I suppose we'll know more once some actual release planning happens for forky, which might take some time.
Are you confusing "386" with 32bit? 686 is the normal 32bit arch. 386 is something from the 1980's right?
When distros mention i386 support they often actually refer to i586 or i686, yes.
True i386 support would mean compatible with the original Intel 386 processor from 1985. The 486 added a few additional instructions in 1989 but things really changed with the Pentium in 1993 - that gave us i586 which is the bare minimum for most modern software today. Much software can still run on regular Pentiums today if compiled for it, but SSE2 optimizations requires at least a Pentium 4 or Core CPUs instead.
I play with retro PCs often and found OpenBSD's i386 target stopped supporting real 386 CPUs after the 4.1 release, and dropped support for i486 somewhat recently in 6.8. It now requires at least a Pentium class CPU, i586, though the arch is still referred to as i386 likely because it's a common proxy for "32-bit".
Debian 3.0 was the last version that ran on the 80386 processor proper, but even as the CPU requirements for the "i386" architecture moved up to the 486, then Pentium, then Pentium II, the name stuck around. Partly from inertia, partly to not break the entire existing mirror infrastructure.
Yes, 386 and 486 stayed relevant throughout the 90s because the price tag for new shiny processors is always higher, and it was not uncommon for customers to favor more or faster RAM/Disk space/graphic card/sound card (that was a thing back then) over better-looking CPU benchmarks.
The Linux kernel also requires at least i486 now. AIUI that decision had to do with smoothing out multicore/SMP support - which is a bit silly because no real 80386 systems in common use are even SMP, let alone multicore. But anyway.
Linux ran fine on 386 chips - that was actually what it was originally developed on. But Intel added a bunch of functionality in the 486, Pentium, and Pentium Pro chips. At some point the powers that be didn't see any value in continuing to support pre-P6 chips anymore.
It was a bit of a strange decision since there were undoubtedly more 386, 486, and Pentium users than some of the platforms Linux continued to support, but that's the choice they made. But they weren't alone. Even NetBSD requires a 486DX or better.
AIUI Debian kept the i386 name for the arch even as their 32 bit requirements evolved.
i386 is the most common term used for 32 bit x86 https://en.m.wikipedia.org/wiki/IA-32
AFAICT this refers to Debian support, the Linux kernel does support 32bit CPUs though only since the original Pentium (excluding some clones).
Well, isn't there an additional year or so of support for old stable? So beyond 2025.
Buster is supported until June 2028.
Not by Debian it isn't.
https://www.debian.org/releases/
Buster has not been supported by Debian for many years.
Buster LTS was EOL last summer. Note that LTS is supported by volunteers via a non-profit, not Debian (though they do a good job).
ELTS is paid support, again not by Debian.
Do look at Debian's wiki for more info on support timeframes, and what LTS and ELTS means.
I was meaning to write Bookworm. Bookworm is the last to support i386 kernel and is supported until June 2028.
Freexian is for-profit, and all the LTS/ELTS contributors are Debian maintainers, and LTS is part of Debian, while ELTS is publicly available too, but in an external archive.
https://wiki.debian.org/LTS https://wiki.debian.org/LTS/Team https://wiki.debian.org/LTS/Extended https://wiki.debian.org/LTS/Funding
Ah, they advertised non-profit at one point, but I see that's changed. That may have been "we seek no profit" not "non-profit entity". Thanks for the info on this point.
Back to LTS:
Debian LTS is not handled by the Debian Security and Release teams, but by a separate group of volunteers and companies interested in making it a success.
To the point, Freexian is 100% not Debian, not "part" of Debian, it merely uses Debian's infra gratis for LTS. This does not detract from the good work they do, but we must also not confuse a private company, and its goals, with Debian and its goals.
LTS tries its best, but only supports what it can. Not its fault. Thus they do give preference to packages which are more widely used, and which they have received donations for.
So wildly popular things such as apache2, mariadb, and so on are very much going to be handled. Some rare package which has 400 users worldwide? Not so much.
LTS will very much take patches and any help, but that still ties in to the number of users. If a packages has 400 users worldwide, and most have moved on to the next release? Well, I hope you see my point.
(I've moved customers off of LTS for using rare packages, whilst reassuring them that LAMP servers are very much supported due to this. Popularity counts here, due to efforts of volunteers and externals.)
--
ELTS only supports a further subset of packages. It's not "full" support. I think one would be exceptionally unwise to use it, for say a desktop. That is, unless they were paying for support and had obtained a list of all packages supported.
--
https://www.freexian.com/lts/extended/docs/debian-10-support...
"Note that when you request a quote, we send you back a list of packages that are not supported or that have limitations in their support so that you can take an informed decision."
Yes, I know that page has a git repo and so on for some support information.
But my points are; not the full distro is supported, you have to track this yourself, you need to be diligent, and even so you need to be sure you're not running rarer packages.
Once again, I do want to reiterate, these are both excellent programs. They do a good job, they're dedicated, but we must be aware of the limitations here.
An example being the differences between security support for main, non-free, contrib in stable Debian:
https://www.debian.org/security/faq#contrib
As you can see, there is no actual guaranteed security support for contrib and non-free. The reasons are logical, however, users need to be aware of the nuance here.
Just as they need to be aware of the nuance of LTS and ELTS.
For example, all of my server installs have non-free, non-free-firmware and contrib blocked via pinning in preferences.d, with only specific absolutely required packages then allowed back in.
(For example I may allow command line apps, but not anything network connected, and only with a once over of functionality and SUID bits and other such things)
--
Really, I see LTS as a crutch that normal users should never use. I suggest we collectively not encourage Desktop users (for example) to use LTS.
The LTS/Freexian people are all Debian members/contributors, so I would not say "Freexian is 100% not Debian" is correct. Basically, some Debian folks got together and started a company to get funding to do an LTS, and also offer other paid services.
At least for bullseye, the LTS team supposedly support all packages, except for games and a few other packages. Its trivial to find out which packages aren't supported too, just run a command, no need to email anyone.
https://wiki.debian.org/LTS/Bullseye https://wiki.debian.org/LTS/Using#Check_for_unsupported_pack... https://salsa.debian.org/debian/debian-security-support/-/bl...
Agreed on the rest, although do note LTS contributors are paid, the security team probably aren't (although some are).
I think in practice, when contrib/non-free stuff has security updates from upstreams, Debian does get updates in stable/LTS. For example the Intel microcode, or WiFi firmware.
I too feel like Debian having LTS is a waste of time, people should be able to upgrade to the next stable within the one year of regular security support for oldstable.
BTW, Ubuntu security support has a similar issue; main is supported, universe is not.
Thanks, I assume some of that is sort of extended, asterisked support though?
Games are unsupported, and a bunch of other packages:
https://wiki.debian.org/LTS/Bullseye https://wiki.debian.org/LTS/Using#Check_for_unsupported_pack...
OpenBSD requires at least a Pentium these days.
Yep! I did some testing and found 6.8 was the final version to support 486 CPUs. The reason is the OpenBSD project switched to clang for the compiler in 6.9, and in the process became dependent on Pentium instructions.
However, that doesn't stop one from installing a Pentium Overdrive in an old Socket 3 board and running the latest release. ;)
Man, I thought I was behind using a P3 in like 2007 lol. You can get something 100x faster for $1 :D
Hahah yeah, it's just a for fun/hobby PC. It's pretty beefy for what it is - 1.3GHz, 512MB RAM, 512GB SSD on a SATA card, fast AGP card. I can run a Debian desktop at 1080p on here.
Debian (and many other distributors of compiled binaries) uses "i386" to refer to all 32-bit x86 processors, including the Pentium 3.
You can still use sysvinit, I've already tested servers and desktop builds.
From my build box:
There is a weird depends you cannot get around without simultaneously removing and installing in parallel. A Debian bug highlighted the above, with a "-" for systemd-sysv- systemd- as a fix, along with allow remove essential.After this fix, sysvinit builds with debootstrap were almost identical as to bookworm. This includes for desktops.
As per with bookworm through buster, you'll still need something like this too:
I can credit you, in version 1.42, with more than a hyperlink to random posts on Hacker News if you have something better elsewhere on the WWW. (-:
* https://jdebp.uk/Softwares/nosh/guide/services/systemd-login...
Wait, sysvinit on debian 13 truly practically works?? as in, one can remove systemd and have a working server OS with sysv init??
Yes, MX Linux will be doing this (as a separate ISO, until now they're been able to provide a single ISO that lets you choose between systemd and sysvinit at every boot):
https://mxlinux.org/blog/changes-coming-with-mx-25/
Yes.
I run a full desktop too, without it. Multiple variants.
I don't use gnome's Desktop Environment though (although I do run gtk/gnome software), so cannot comment on that.
Why would you want to do this?
To avoid systemd, presumably. Although one could also just switch to Devuan at that point.
choice.
Thank you for sharing this. I'm inclined to adopt it in my lxc containers, at least.
I've been living with sysvinit up until Debian 11. Then it became unusable with lxc containers :(, so I had to bite the bullet. But for the basic system it indeed works.
A new APT sources format "debian.sources" is announced with trixie. The now older "sources.list" format is still supported, but is likely to be deprecated in a future Debian release.
See below:
- https://wiki.debian.org/SourcesList#APT_sources_format- https://www.debian.org/releases/trixie/release-notes/upgradi...
"apt modernize-sources" command can be used to simulate and replace ".list" files with the new ".sources" format.
> apt modernize-sources
Oh nifty, I hand converted all mine a couple years back. It would have been nice to have that then (or know about it?). I do really like the new deb822 format, having the gpg key inline is nice. I do hope that once this is out there the folks with custom public apt repos will start giving out .sources files directly. Should be more straightforward than all the apt-key junk one used to have to do (especially when a key rotated).
Same. It took me a little bit to get used to it; my initial snap judgment was “this will be more annoying to create via scripting,” but then Ansible added deb822_repository [0] in 2.15 (shortly before Bookworm was released), and then it was no longer a concern.
[0]: https://docs.ansible.com/ansible/latest/collections/ansible/...
DEB822 was available from at least Buster [0]. I think Bullseye was the first release I used it in.
[0]: https://manpages.debian.org/buster/apt/sources.list.5.en.htm...
Ooh, that's nice. Especially nice that it lets you specify both Suites: and Components: in the same stanza, so you don't have to repeat the rest of the line to add -updates and -backports suites.
It was announced in 2015 with apt 1.1 which was a major change of configuration, different things from it are being enforced each release since.
They only mentioned it briefly, and not by number, but this release includes 95%+ bit-for-bit reproducibility on AMD64, ARM64, and RISC-V across more than 30,000 packages (92% mean across all architectures).
Congratulations to the team--phenomenal work!
https://reproduce.debian.net/
Is there a tool on a given debian trixie system to know what installed packages are not currently reproducible?
Alternative to parsing the reproduce web site :)
Yes! From the site: sudo apt install debian-repro-status; debian-repro-status
Could you help me understand why the remaining 5% is not bit-for-bit reproducible? For example... if you download a tar of sources pinned to a version, and you run `./configure` and `make` in some kind of container and it doesn't embed some kind of timestamp... why are 95% reproducible and some aren't? Would like to learn/understand.
Hashtables keyed off the address of objects would be an example.
On multiple runs, malloc gives out different addresses (thanks to threads or security concerns) which means things end up in different slots in the table. Then you iterate through it in memory order and you're seeing objects in non-deterministic order, which you do things with.
Embedding file paths / timestamps / git shas and similar was popular for a while too and unhelpful for reproducible builds.
I see that systemd is still doing this thing where they are trying to strong-arm all Linux distros into arbitrary stuff that someone decided is the only right way to do something:
> 5.2.2. systemd message: System is tainted: unmerged-bin systemd upstream, since version 256, considers systems having separate /usr/bin and /usr/sbin directories noteworthy. At startup systemd emits a message to record this fact: System is tainted: unmerged-bin. It is recommended to ignore this message. Merging these directories manually is unsupported and will break future upgrades. Further details can be found in bug #1085370.
No option to disable this either, per discussion in https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1085370
You may not like it, but “arbitrary” isn’t a fair description; there’s reasoning behind it that is over 10 years old:
http://0pointer.net/blog/projects/stateless.html
https://www.freedesktop.org/wiki/Software/systemd/TheCaseFor...
That said, my knee-jerk is also that this is about strong-arming distros. Which leaves a bad taste in my mouth. I’d be interested to hear other viewpoints though.
Note that this is about a different thing - Debian has also merged its /bin into /usr/bin, but now systemd also wants /usr/bin and /usr/sbin to be merged.
Oh thanks, I totally missed that distinctions. You’re right.
My first thought is to wonder: why would systemd care about this then? Any idea?
My second thought is: with a merged /usr, where /usr/{bin,sbin} are on the same filesystem, what’s the benefit of even having a distinct sbin? My understanding is that sbin has historically been useful to have statically linked binaries that can be made available early during boot without needing various lib dirs mounted. But that need seems obviated by a unified /usr…
> with a merged /usr, where /usr/{bin,sbin} are on the same filesystem, what’s the benefit of even having a distinct sbin?
There really isn't any, the sysmlink being for compat. Early boot is nowadays taken over by an initrd/initramfs. Some say what you say sbin was for were actually the root /bin (early boot) and /usr/bin (elsewhere) but it was all just so arbitrary and no-one could actually agree on what is for what actually.
Some say sbin is for static binaries, others for superuser binaries and either destinction really doesn't make sense nowadays as earlyboot is initrd and a lot of "superuser binaries" actually also are fine to use unprivlidged in certain scenarios (e.g. `mkfs.*` for making disk images) so having them in a separate dir is just arbitrary (e.g. `cat` can also fail if I cat the shadow file).
>No option to disable this either, per discussion in https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1085370
The discussion in that bug is that the Debian maintainer (and upstream dev) is open to an upstream patch to add such an option.
> Debian maintainer (and upstream dev) is open to an upstream patch to add such an option
Wild interpretation right here.
There are only 2 realistic choices: Leave it as is or patch out the warning message in the Debian package
Debian maintainer is clearly deflecting the responsibility here because everyone knows very well that upstream wouldn't accept such a patch.
As it's already explained in the bug report, since Debian has no plan to do that migration in the near future, aforementioned warning isn't only useless and annoying, it's also potentially harmful, thus the correct action would be to remove it downstream like they did it in xscreensaver (https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=819703#84)
But then they'd face the wrath of Lennart, so the only choice left is ignoring the report.
Perhaps we need to have a concept of separation of upstream and distros.
Why do you think this is an attempt at a persuasion tactic? Taint flags in this context just means something that might be relevant to debugging. Which this condition might be, if people in the future are unfamiliar with a potentially anachronistic split between /usr/bin and /usr/sbin. The debug message isn't there to judge the morality of your configuration. It's actively improving your ability to continue to support both ways, by properly indicating what style of system it is for troubleshooting purposes.
> Why do you think this is an attempt at a persuasion tactic
Because they said so:
> As part of that we sometimes adopt schemes that were previously used by only one of the distributions and push it to a level where it's the default of systemd, trying to gently push everybody towards the same set of basic configuration [1]
1. https://0pointer.de/blog/projects/the-biggest-myths.html
This was written 11 years before that debug log was added. Extremely loose connection. Things do actually change over the course of 11 years, believe it or not.
It's their modus operandi and it haven't changed, believe or not.
> it is definitely our intention to gently push the distributions in the same direction so that they stop supporting deviating solutions for these things where there's really no point at all in doing so. [1]
> Distros may deviate from this by patching this downstream, but by shipping this as secure default I do hope to gently push everybody in the same direction.
> start pushing people gently to define GPT partition type UUIDs for missing archs > Let's generate a single gcc `#warning` message asking people to define partition type UUIDs for their architectures if they are missing. [3]
[1] https://lists.freedesktop.org/archives/systemd-devel/2010-Se...
[2] https://lists.freedesktop.org/archives/systemd-devel/2011-Ma...
[3] https://github.com/systemd/systemd/commit/d42e4fa258701ff5e0...
How and why is this relevant to debugging?
Also, like it or not, the word "tainted" bears a negative connotation. It's hard to imagine that it was chosen arbitrarily.
Lovely.
Debian choosing systemd (not that it's a new decision) is the reason I'll be switching my Proxmox to FreeBSD/bhyve (FreeBSD has great ZFS support btw).
Once I get the hypervisor systemd-free (no systemd on FreeBSD), I can then install a minimal distro in a VM mean to do containerization (like, say, the Talos Linux distro for K8s, that only has a few executables and they're all immutable) and then I can run containers that, by design, have something that is precisely not systemd as PID1.
So life is good: there's a systemd-free world at the end of the tunnel.
> Debian choosing systemd (not that it's a new decision) is the reason I'll be switching my Proxmox to FreeBSD/bhyve
Did you consider Devuan? Or is this just taking one annoyance as motivation to fix others at the same time?
It can be a little hard to navigate to so the .torrent links for x86-64 are
Minimal: https://cdimage.debian.org/debian-cd/current/amd64/bt-cd/deb...
Full: https://cdimage.debian.org/debian-cd/current/amd64/bt-dvd/de...
Note that you probably don't need the DVD ("full") image. Most users should use the "minimal" netinstall CD and download packages at install time.
I downloaded both because I intend to seed, but yes, if you have a fast internet connection then minimal is perfect... but if you are on a crappy third-world connection where it might take all day to get the image, it's nice to have it all in place when you are ready to install.
Agreed, but for laptops it’s nice to keep a copy of the DVD iso on disk and in your apt sources so that you can install stuff offline.
It baffles me I keep searching for these files every time I want to download Debian. And it’s been like that for over a decade at least.
They are linked from the "Other downloads" page linked from the front page.
Yeah, I think I remember that, but each time I’m puzzled and not sure. It feels like they hid that deliberately. While Arch is the opposite, it’s so easy to get the torrent file.
Yeah, it was deliberate, to make it easy for most people, who will only want the netinst download.
Biggest change for me is /tmp behavior. In Debian 13 /tmp become RAM-disk by default (instead of files on the file system) and uses up to 50% of available ram. But as expected of Debian the release notes included an easy fix to restore normal /tmp behavior for people and applications that place many small or large files there.
https://www.debian.org/releases/trixie/release-notes/issues....
>"You can return to /tmp being a regular directory by running systemctl mask tmp.mount as root and rebooting."
I kind of wish the distros had decided on a new /tmpfs (or /tmp/tmpfs, etc) directory for applications to opt-in to using ram-disk rather than replacing /tmp and having to opt-out.
This was discussed a ton in debian-devel. First, the tmpfs doesn't take much space already, and /tmp became a folder where persistence should not be expected over the years.
The problem with /tmp was many people and apps used it as an inter-user communication medium and expected persistency there, so it created both security problems and wasted disk space over time.
Since not many packaged apps used the /tmp like that and used the folder the way it should be used, the change was made.
I'm running Debian testing on one of my systems, and the change created no ill effects whatsoever. Not eating SSD write cycles can be considered a plus, even.
However, as I also noted in the relevant thread, the approach might have a couple of downsides in some scenarios.
If you have the time and the desire, discussion starts at https://lists.debian.org/debian-devel/2024/05/msg00014.html
Infinite scroll length on terminals can chew through /tmp, and systems misbehave strangely when they're out of /tmp.
What terminal emulators default to infinite backscrolling? Genuinely curious here. I've used xterm with a scroll buffer of 512 lines for three decades now and outside of large builds I can't imagine wanting more than that.
Also watch out for surprise file deletes in /tmp and /var/tmp at 10 and 30 days.
This too can be turned off.
Isn't this the feature of /tmp? I set my default download location in Firefox to /tmp exactly for this reason, so that all the junk gets automatically removed after some time. Also, whenever I need a temporary Python script or test a package, I create a venv under /tmp.
On boot has been the standard for a long time and is still the most common. I am personally surprised to hear that now Debian and some distros do it via various automated ways at time intervals.
It's a systemd thing, see `man systemd-tmpfiles`.
It was available as an option before that: https://manpages.ubuntu.com/manpages/xenial/man5/rcS.5.html
I think that superkuh's point is that it is not a systemd thing. Cleaning up /tmp by deleting old files has been around since before systemd was invented. Since before Linux was invented, even.
Yes but in Debian this was not a default until now.
Before, /tmp was wiped on reboot. /var/tmp was not.
Neither were cleaned otherwise.
So for Debian, this is a systemd thing. And it was pushed by a systemd maintainer, who is also a Debian developer.
I have zero interest in Debian "being brought inline" with other distros, because other distros should be coming inline with Debian.
Outwith the world where only Debian and other Linux distributions exist, though, this isn't really systemd doing bad things to poor old Debian. To a wider world it is Debian and other Linux distributions exceedingly slowly reinventing things from Unix.
AIX had /usr/sbin/skulker from at least as early as version 1.2.
For AT&T System 5, Fielder and Hunter were popularizing a similar utility named rmtrash that is run nightly.
Fielder and Hunter exemplify traditional Unix thinking in their 1986 book on Unix system administration, which includes rmtrash:
> By putting all temporary files in one or two directories, it is easy to clean them all out at regular intervals. For this reason, it's a good idea to encourage users to use /tmp for all files they need only a short while.
skulker is not in the old comp.unix.aix Usenet FAQ document, but given the number of times "That's what skulker does, and you should be running it daily." seems to have been the answer over the years, it probably should have been. (-:
To old hands, this is not a systemd novelty at all. The novelty, to old hands, is the idea of doing this not using a script. I remember the war stories from the 1990s and turn of the 21st century when users did things like include LF in the names of temporary files, and administrators suddenly learned the utility of find -print0 and xargs -0 or find -exec rm {} +. Stéphane Chazelas for one has had a lot to say on the subject over the years.
That said, mtree (from 1989) already existed at the point that systemd-tmpfiles was invented, and already had the idea of working from a specification file with names and owners and permissions and whatnot. It is surprising that no-one ever apparently tried mtree for completely wiping /tmp. The BSDs are using find to this day in the several places with they auto-delete stuff in /tmp including the daily periodic, although they did spot that their own find had a -delete option in 1999. (-: Hell, even I am still using find.
https://github.com/freebsd/freebsd-src/blob/main/usr.sbin/pe...
I can’t fathom why anyone would be surprised that a directory named “tmp” is ephemeral.
Then I think you just don't have much imagination. I have recovered files in /tmp after turning off a machine by booting it back up in single-user mode and accessing the data before it would be cleared in during bootup. Given that "turning off a machine" can also mean "the machine lost power", I can definitely see why people would be surprised by this change.
It could be the 20+ years of using linux where it was standard for it to only have files there deleted on boot. Temp in this case means it's temporary for that boot. Recent redefinitions are surprising unless you've only come to linux recently or from unix 40 years ago.
I am going to assume you are used to tens of days wipings of /tmp. Imagine suddenly it was decided that the wiping needed to happen every tens of seconds for security reasons and to minimize ram usage. Do you think this would impact your workflows that used /tmp? This is 100% analogous to the current situation. Yes, tmp is ephemeral, but on what timescale?
No, I misunderstood the comment I was replying to, and thought they were assuming they could use /tmp as an un-curated dumping ground that was never deleted.
I also am used to it being wiped on boot.
A total of seven architectures are officially supported for "trixie":
It's good to see RISC-V becoming a first-class citizen, despite the general lack of hardware using it at the moment.I do wonder, where are PowerPC and IBM System z being used these days? Are there modern Linux systems being deployed with something other than amd64, arm64, and (soon?) riscv64?
Mainframes are still holding on in use cases where a single server having continuous uptime is vital. They're designed to have uptime measured in decades, so even components like the processors and main memory have hot spares available and can be hot-swapped without interrupting the OS or running services. They also have continually running system monitoring and diagnostics at the hardware level (not running as an OS service) that will alert both the owner and IBM if they detect some sort of hardware fault. IBM has supported Linux as a first-class OS option for their mainframes since the early 2000s.
From a developer perspective, s390x is also the last active big-endian architecture (I guess there's SPARC as well, but that's on life support and Oracle doesn't care about anyone running anything but Solaris on it), so it's useful for picking up endianness bugs.
Another interesting thing is that the only two 32-bit architectures left supported are armel and armhf. Debian has already announced that this will be the last release that supports armel (https://www.debian.org/releases/trixie/release-notes/issues....), so I guess it'll be a matter of time before they drop 32-bit support altogether.
Yeah, mainframes are all about cases where the uptime is critical. Most modern systems are good with offering 99.9% or 99.99% reliability, with the understanding that trying to offer more than that just gets more and more expensive. Well... spending huge amounts of money on mainframes is one of the strategies to get to reliability numbers like 99.9999999%.
Source: https://itic-corp.com/itic-2023-reliability-survey-ibm-z-res...
The legitimate usage is typically for workloads like relational databases (which are anyway single-machine architectures) that experience heavy load 24/7, part of fragile architectures that cannot tolerate downtime (remaining fragile as such because the original source code has been lost etc.), where "the system is down" causes tens of thousands if not hundreds of thousands of people to stop work.
Both Power and z are many billion dollar businesses each. Banking and other high finance is the stronghold for both. IBM still seems proud of z, Power seems merely tolerated these days which is a shame because it is a nice ISA and the systems are very nice too.
IBM puts a lot of work and money into making sure open source stuff runs properly on those two, even if they aren't that popular
them being kept by major distros is therefore not as "natural" as other architectures
For Debian s390x (IBM Z), they only employ two people to work on the port, which isn't really enough. I think ppc64el has even less people.
> I do wonder, where are PowerPC and IBM System z being used these days?
IBM.
And they own redhat, so I imagine they put a lot of time and money into making the kernel work.
Why Debian in particular, not sure.
Very important to note the lack of x86 (32bit x86) support.
The end of an era.
I have used Debian starting sometime around slink. I still type "apt-get ..." and it still works reasonably well. There have definitely been hiccups in upgrades over the last 25+ years but the amount of time/effort dealing with those is almost nothing in comparison to other linux and non-OSS systems I've dealt with over the same span of time. My only regret is not contributing more to the community.
The thing I like most about Debian is that you need to know at least a little about what is going on to use it. For me, it does a good job of following "as simple as possible and no simpler."
Which one was slink?
My first Debian install was in 1996. I had no real idea what I was doing, but it was amazing to me that I could remote-display windows from machines across campus, and it was alien compared to the windows 3.x/95 I was used to at that point. There was no apt at that point, or none that I was aware of, and adding new stuff was painful.
I started using debian preferentially as my workstation/desktop OS in about 2005, and was installing it on embedded systems (linksys nslu2) to make micro servers by … etch I think it was.
By 2008 I was at IBM and they allowed a choice of windows or redhat on your laptop, and if you were adventurous there was experimental support for Ubuntu which might work on Debian. I made it work and discovered that among 330k people there were 22 of us running it!
Always loved it, it always just made more sense than other distros somehow. My daily driver is a Mac now, but I still have a few Debian machines around.
As an owner of two i386 systems (both netbooks built around Intel's Atom N270), that run Debian, I am a little sad. I understand the reasoning, and I won't deny it is a very niche platform by now. But I had hoped Debian, with a history of supporting a wide range of platforms, would keep i386 going for a while longer.
Fortunately, bookworm will continue to receive updates for almost 3 years, so I am not in a hurry to look for a new OS for these relics. OpenBSD looks like the natural successor, but I am not sure if the wifi chips are supported. (And who knows how long these netbooks will continue to work, they were built in 2008 and 2009, so they've had a long life already.)
EDIT: Hooray, thanks to everyone who made this possible, is what I meant to say.
OpenBSD runs perfectly fine. Atom netbook, n270, 1GB of RAM, cwm+git dillo (plus DPI plugins), mpv+yt-dlp.
My ~/.config/mpv/config:
My ~/yt-dlp.conf For the rest, I use streamlink from virtualenv (I do the same with yt-dlp) with a wrapper at $HOME/bin:yt-dlp wrapper
streamlink wrapper To install streamlink The same with yt-dlp: On the rest, I use mutt+msmtp+mbsync, slrn, sfeed, lynx/links, mocp, mupdf for PDF/CBZ/EPUB, nsxiv for images, tut for Mastodon and Emacs just for Telegram (I installed tdlib from OpenBSD packages and then I installed Telega from MELPA).Overall it's a really fast machine. CWM+XTerm+Tmux it's my main environment. I have some SSH connection open to somewhere else at the 3rd tag (virtual desktop), and the 2nd one for Dillo.
Thank you very much!
Alpine supports i686, I see no current deprecation plans. This may change in the next three years though, who knows.
Out of curiosity, what do you use these netbooks for?
One sits in my bathroom so I can browse random Wikipedia articles while I'm, uh, busy. The other one sits on my nightstand and plays audiobooks/podcasts when I'm going to sleep.
So nothing critical. But something they are still good at, and being very small makes them a natural fit for these use cases.
Curious, why not use your phone for both these use cases? Seems like it would be even more convenient
I can't speak for the other poster, but I like the idea a lot. Having tools with specific purposes means I can avoid using my phone for everything. No matter what games I play to remove notifications/interruptions/etc. it's always a distraction and easy to be distracted from whatever I originally intended to use the phone for.
I do use the phone for audible, but I started both uses before I had a smart phone (I was very late to the game), and I am a creature of habit. Plus the netbook has a bigger display, more storage, and a real keyboard (again, creature of habit).
antiX will be creating a Trixie-based 32-bit ISO. There's also Void, Alpine and Slackware (at least).
For those worrying about the NIC change with systemd, this comes from the release doc:
https://www.debian.org/releases/trixie/release-notes/issues....
Here's a one-liner, excluding a bond interface and lo. Gives a nice list of pre and post change. The doc's logic is that after you've upgraded to trixie, and before reboot, you're running enough of systemd to see what it will name interfaces after reboot.So far I have not had an interface change due to upgrade, so I cannot say that the above does detect it.
Hopefully the last breaking change.
enoX should always stay stable, as it's the BIOS (in some ACPI table) telling that this device/port has this ID.
ensX means the NIC in PCIe slot X, but in your PCIe tree you can have PCIe bridges, so technically you could have multiple NIC in the same slot (what the BIOS declare as a slot), so there was a lot of breaking NIC naming changes over the years in systemd to figure out the right heuristics that are safe, enabling/disabling slot naming if there is a PCIe bridge, but just in some cases.
Also for historical reasons the PCIe slot number was read indirectly leading to some conflicts in some cases (this was fixed in systemd 257)
>Hopefully the last breaking change.
Every year's cope with systemd.
Do you happen to know if this change can affect people who have disabled systemd's Predictable* Network Interface Names before upgrading to Trixie?
*haha
You can easily keep the current naming behavior with the 'net.naming_scheme=' kargs
I already have systems that use net.ifnames=0. My question is about whether this new behavior can affect them.
It'll not, the new behavior is just a new naming scheme, and you just choose not to use any, which is totally fine if you have a single NIC.
Debian's signature feature (upgrade from stable to stable under 15 minutes) shines here too.
My first system migrated in less than 10 minutes, incl. package downloads and reboot. It's not a beast either. N100 mini PC connected to a ~50mbps network.
What was the process for this? Does it involve manually modifying the sources.lists files, or is there a single command to instigate the upgrade?
There are two main ways:
If your sources file references the release name (e.g.:bookworm), you change them to trixie, then “apt update && apt dist-upgrade”.
or,
If your sources file directly reference distro-suites (e.g.: stable), you just “apt update && apt dist-upgrade” since stable is now pointing to trixie.
In the first reboot, you run “apt autopurge” to remove packages which are not needed anymore.
…and you’re done.
Is this really a signature feature when rolling release distros are common?
Yes, esp. if you are using Debian at infra and in production. I upgraded three servers in the morning while sipping my tea. Nobody noticed anything.
For RedHat family, this is nigh impossible, requiring hours of planning, downtimes, etc., and even then, it's not guaranteed to work.
If you prefer a rolling release, and won't use it on a server, Debian Testing got you covered (unless you are in a security sensitive environment). My systems are well-isolated from outside, so desktop systems can run Testing without issues.
Servers and exposed systems always run stable with security updates installed automatically, though.
I've been on Debian forever and love it to bits. But in an act which can only be described as batshit insanity, they have chosen to patch Python's pip3 in a manner which breaks the --prefix option. On Debian
will install into /usr/local/local, so one has to use the prefix /usr. The same command on, say, OpenSuSE will install into /usr and break your system. Barking mad.https://sources.debian.org/src/python3.7/3.7.3-2+deb10u3/deb...
> python3.7
Certainly a terrible UX, but the motivation is clear: they're trying to get PEP 668 protections for older versions.
Virtual environments work a lot better anyway, honestly. (With a properly crafted `pyvenv.cfg`, it should be possible to convince Python that your /usr/local is a virtual environment, but I can't be sure offhand if there are any serious negative consequences of that.)
python3.7 is from June 27th, 2018
https://www.python.org/downloads/release/python-370/
Why is that the default?
I don't think it is? Or at least, it only was on some particular older Debian release.
I think that was the version the patch was introduced, it's certainly in-place on Trixie (which has 3.13).
Oh. If it's still there, I guess nobody's given them any impetus to revisit it...
Lots of debian love in this thread and it's great to see. If you're so inclined I encourage you to donate to Debian. We're all better off the more support goes to an ecosystem and operation like Debian.
https://www.debian.org/donations
Not affiliated, just a happy user for a long, long time.
Lots of other ways for individuals, companies and other organisations to help too:
https://www.debian.org/intro/help
I have been using Debian Trixie for a few months in testing now, I can attest that its a great, stable operating system. Definitely better than Ubuntu in terms of user experience.
The only complaint on a fresh install is that Cinnamon seems to use a ton of CPU when there's a little moving thingy anywhere on the screen (a browser tab that has a loading icon in the tab list is sufficient). This is most noticeable when you have a VM without graphics acceleration (don't ask why in the world my job requires that). Graphics without acceleration is always heavy, but this is an extra process doing whatever on top of the actual load
Then my private laptop has had a bunch of graphic issues after upgrading to 13 (it manifests differently in a lot of applications and it changes when you pick a different desktop theme, not even sure how to describe it). The new pipewire (pulseaudio replacement, idk why that needed replacing) does not work properly when the CPU is busy (so I currently play games without game sounds or music in the background). The latter then also sometimes (1 in 5 times maybe?) crashes when resuming from suspend, but instead of dying, spams systemd which diligently stores it all in a shitty binary file (that you can't selectively prune), runs completely out of disk space, and breaks various things on the rest of the system until you restart the pipewire process and purge any and all logs (remember, no selective pruning)... Tried various things I found in web searches and threw an LLM at it as well, but no dice. I assume these issues are from it not being a fresh install, so no blame/complaint here really, just annoying and I haven't had these issues when doing previous upgrades. Not yet sure how to resolve, perhaps I'll end up doing a completely new install and seeing what configs I can port until issues start showing up
Surely these things are not a Debian-specific issue, but I haven't noticed something like that with either 11 or 12
Edit: oh yeah, and the /tmp(fs) counter is at 1 so far. I wonder how many times I'll have run out of RAM by Debian 14, by forgetting I can't just dump temporary files into /tmp anymore without estimating the size correctly beforehand
I've been testing Cinnamon on my Ubuntu machine recently. I was close to deciding to switch from Gnome but I keep running into an issue whenever I have long running tasks in Blender.
In Gnome, Blender becomes unresponsive but everything else is still usuable. In Cinnamon, the entire system becomes unresponsive.
> "trixie" includes numerous updated software packages (over 63% of all packages from the previous release)
Wow, I'm amazed a third of packages haven't seen an update in, ehm checks
> After 2 years, 1 month, and 30 days of development, the Debian project is proud to present its new stable version
I'm a fan of old software myself, in the sense that I find it cool to see F-Droid having a (usually tiny) package that is over 10 years old but it does exactly what I want with no bugs and it works perfectly on Android 10. I wonder if those 30% more commonly fall in the "it's fine as it is" category or in the "no maintainers available" category
> Users running i386 systems should not upgrade to trixie. Instead, Debian recommends either reinstalling them as amd64, where possible, or retiring the hardware.
What I did is switch to NetBSD.
In the abstract I'm a big fan of supporting me old machines forever, but I have to ask out of curiosity - what hardware is practical to run these days and only has a 32-bit processor?
I have a few old PCs (towers) here which don't support amd64 mostly Pentium 4-based.
They all still have DVD reader drives and are nice for ripping CDs. Despite the fact that the drives are nearing 20 years of age (machines are from ~2005) they still perform better than most “new” external drives. Of course one could also move the drives to a newer machine but many of them use the IDE connector which is not commonly found on modern systems. Also, modern cases typically don't account for (multiple) 5.25" drives.
The other use case is to flash microcontrollers. When fiddeling around with electronics there is always a risk of a short circuit or other error that could in worst case kill the attached PC's mainboard. I feel much safer attaching my self-built electronics to an old machine than to my amd64 workstation.
Due to their age, I think the old machines may not live much longer -- I fear not even 10 more years, some of my old 32-bit laptops have already failed. Hence even for me it does not make sense to try keeping up the software support. Maybe I switch them to a BSD or other Linux distribution if they live long enough but for now the machines run OK with Debian Bookworm (newly oldstable), too.
I tried to repurpose an old laptop I had lying around as a "lie on the couch and surf the web or watch youtube" machine. It was one of the last 32-bit only cpus (pentium m), so I installed Debian bookworm (12) on it. Unfortunately it turned out it couldn't even play youtube videos at 144p without stuttering. So I E-wasted the machine.
I suppose as some kind of headless home server it could still have been useful. OTOH for something that runs 24/7 a RPi would use a fraction of the electricity and still be a lot more powerful.
So yes, beyond nostalgia and some embedded/industrial usecases, it's hard to see a use for a 32-bit only PC these days.
In my case I have a couple of first gen 32-bit Atom netbooks that I use regularly for the same things I've always used them for. The hardware still work just fine so I see no reason to replace them.
The difference between Debian and Ubuntu is decreasing with each release recently. I was pleasantly surprised that Debian recognized all hardware components in my laptop released one year ago out of the box.
Hardware support is good and UI is great! It feels snappier than Ubuntu, may be due to lack of snap and fewer services and applications installed by default.
I just upgraded a mini pc with no real issues. Main steps were:
I love Debian and have a tremendous amount of respect for the people who work on the project. I no longer use Debian, but I think it's vitally important to have an anchor Linux distribution which isn't overly reliant on a for-profit entity and is truly community driven.
Arch isn't truly community driven?
> The overall disk usage for trixie is 403,854,660 kB (403 GB)
What does this mean? If all 69k+ packages are installed, it will take up this much space?
As this also lists lines of code, it sounds more like sources plus packages. Think space that a full mirror (src + generic + arch specific packages) would need.
Indeed, this is the amount of space that a Debian mirror would need to host all Trixie packages. So it's the compressed packages total size, not the space it would take to have all packages installed simultaneously (which also happens to be impossible, because of package conflicts/alternatives and Debian supporting 7+ different architectures).
The 403GB figure represents the total size of all source packages in the Debian archive, not the disk space required for a typical installation which is usually under 10GB for a desktop system.
I've upgraded all my servers and laptops to Debian 13.
Lucky 13 and all... And not a single issue so far. Very happy!
Thanks to the Debian team for putting out yet another high quality, reliable release :)
I'm very sad to see them drop support for 32-bit, since that is the computer on which I have been using Debian for the past 10 years...
Does anyone have any suggestions for a 32-bit distro that's still being updated?
Linux mint technically still does, perhaps because they’re lagging behind Debian. AntiX does but that has its own tradeoffs. With full potential respect for your situation, if you’re not able to obtain 64 bit hardware, there’s nothing that wrong with not upgrading your OS for now
Bookworm will get updates for another year if I have read the Debian Wiki correctly so perhaps there is no need for immediate action as Bookworm becomes oldstable?
In addition to Mint and AntiX there is also Slackware (I use xfce rather than KDE) but adding software outside the (large) base install is not an 'apt-get thing' process.
Salix Linux, based on Slackware, does have package repositories with a reasonable selection of applications. Salix is based on the stable Slackware 15.0 so has nicely aged packages.
Void Linux also has an xfce4 based live i686 iso you can look at and decide to install from.
I also use xfce, my computer is an Acer Aspire One, limited in ... well everything, I did upgrade the storage to an SSD, but it's still an Atom CPU, so whatever is the lightest is the best.
Gentoo still supports it but depending on what you want, compile times with modern compilers on old hardware are a pain.
If you're upgrading, see https://www.debian.org/releases/trixie/release-notes/upgradi....
Looking forward to 13! Debian 12 in combination with Pipewire is my go to daily use professional audio workstation for 2 years. Coming from Windows, there are no more forced updates anymore preventing me from doing my job. This is a releave. It works so good! Linux for professional audio is really an option now! Most high-end converters are connected via MADI or ADAT anyway, so there is no driver problem existent. Drivers are a consumer grade discussion..
I've kind of been using Debian 13 for awhile now (I'm on Unstable) and for me what's impressive is how polished a default Debian installation is these days. With Gnome, you literally can run it as is, no config needed. It just works.
That being said, I like Flatpak, so I installed it (was super easy and Flathub provides instructions), and I added a few Gnome Shell extensions (a Dock so my wife can find apps when she occasionally uses my laptop).
Debian gives you a feeling of ownership of your computer in a way the corporate distros don't, but is still pretty user friendly (unlike Arch).
I'd definitely install Debian Stable on a grandparents' computer.
I used to like Debian when configuring ALSA/OSS, XFree86 and such was a source of nightmares. Thus, debconf as a middle layer mechanism to handle several distinct architectures, setups and hardware was a neccesity. Ditto with Yast2 on SuSE. By 2004-2005... not much.
Even a bare Slackware with KDE and KDEi (and even XFCE) can do tons of work by itself by just adding an user and accepting the default group belonging array by pressing 'up' at the prompt.
Heck, even OpenBSD, minus the volume automount, which can be handled in a breeze with toadd or tray-app in seconds; and if you are smart you can figure DBUS/FDo mount points and integrate then with XFCE/Plasma/Gnome without too much issues (hotplugd can handle device umounting if you set up doas.conf accordinly).
The rest? MESA and X.org will handle most of the graphics stuff. Video and audio drivers are autodetected on almost every GNU and *BSD. Printers are often wireless bound so any assistant with look it up fast and attach it to CUPS.
Still, I can't handle DPKG/APT's slowness, even if there are libre distros as Trisquel with it. If they rebased their distro as a simpler Parabola LTS release with either Mate or LXDE setups, the user experience would be almost the same, but installing packages would happen at a much faster pace.
dpkg/apt in Debian fells slower even compared with dpkg/apt in Ubuntu, not sure why that's the case
Maybe a niche concern, but SDL2 is still in Trixie. The sdl2-compat layer (translating SDL2 APIs to SDL3) is in testing, where SDL2 also exists side-by-side with it and is intended to be used to test and verify that SDL2 apps that use it are actually compatible.
Night-and-day decision-making process compared to Fedora and Arch, which both replaced SDL2 with sdl2-compat, broke a bunch of SDL2 apps because sdl2-compat isn't actually SDL2-compatible yet, and sent everyone to yell at the SDL team about it.
Well Fedora is the equivalent of testing for Red Hat and Arch is Arch.
Even Arch kept SDL2 libs in the repo as a fallback. Fedora jumped straight from SDL2 in the repo to sdl2-compat or nothing. The only way to get consistent SDL2 builds on F42 is by building SDL2 from source.
Debian 13 trixie includes numerous updated software packages (over 63% of all packages from the previous release)
I’m not familiar with the metric definition they use, but I’d be worried if close to 100% of the packages they included in bookworm hadn’t been updated in the roughly 2 years between releases.
I use Debian for most of my servers, so I’m sure there is a valid explanation of that phrase.
I don't know why you think it would be different. Are you concerned about security updates? That's not part of the metric, as far as I can see.
And even if it was?
If you look at the number of packages in Debian, only a small portion have CVEs. There are nearly 30k package sources, and an output of 60k binary packages.
Yet we only get a few security updates weekly.
Another example? Both trixie and bookworm use the same firefox ESR (extended release) version. Both will get updated when firefox forces everyone to the next ESR.
Beyond that, some packages are docs. Some are 'glue' packages, eg scripts to manage Debian. These may not change between releases.
Lastly, Debian actually maintains an enormous number of upstream orphaned packages. In those cases, the version number is the same (sometimes), but with security updates slapped on if required.
From my perspective, outside of timely and quick security updates, I have zero desire for a lot of churn. Why would I? Churn means work. Churn means changed stability.
We get plenty of fun and churn from kernel, and driver related changes (X, Wayland, audio/nic, etc), and desktop apps. And of course from anything running forward, with scissors, like network connected joy.
> I’d be worried if close to 100% of the packages they included in bookworm hadn’t been updated in the roughly 2 years between releases.
Code doesn't "go bad" and not everything is affected by ecosystem churn and CVEs.
An established package not having updates for 2y is not in and of itself problematic.
It's not uncommon for small software packages to go years between updates - either because they're a simple utility that's feature-complete and rarely needs bug fixes, or because they're data files (e.g. packages of icons or fonts) which might not need to change at all.
Debian stable is just that - unchanging between major Debian versions. They do however push security updates when necessary, so you're not missing out on those
Any chance you know how they manage that? Surely not every package in the repos is supported for the entire 2 year cycle, so if a vuln comes out after a major refactor, it’s surely not easy to backport the patch.
They auto-import CVE feeds into the security tracker, file bugs for Debian maintainers to fix the issues, curate the tracking data, coordinate with upstreams and other distros to get fixes and so on. Some more on the team web page.
https://security-tracker.debian.org/ https://security-team.debian.org/
Theres some information here they've put out: https://www.debian.org/security/faq
And yeah it must be an incredible amount of work to stay on top of all this
If upstream makes no releases in that time, then there'll be no upgrades.
Plasma 6.3 - I can finally ditch kde neon.
Not if you want to remain on new Plasma, you can expect Debian to lag several minor versions behind.
I've found it pretty easy though to use some KDE components built from source on top of the standard Debian packages. Build with kdesrc-build, then have those binaries linked to from your ~/bin and you're set. It might get difficult if you want to rebuild some key components like plasmashell itself but I've been using locally built versions of Kate and Konsole without issue.
> you can expect Debian to lag several minor versions behind.
Not necessarily forever, though. Bookworm got minor Plasma updates, so I wouldn't be surprised if Trixie does as well.
Bookworm stayed on Plasma 5.27.5 when Plasma shipped bugfix releases up to 5.27.12. Debian may have cherry-picked a handful of patches from there, but that's a lot of bugfixes missed on a release that was already super old.
Even at this point, Plasma 6.4 has been out for almost two months and 6.3 will not get any more updates ever. While everyone else is upgrading, Debian is going to be stuck on an already unsupported version for another two years or however much.
Debian is great for what it is, but you better hope you don't run into issues with your desktop environment because they will not be addressed.
There are options if you're willing to put in the work and run the risk of breaking things.
You can always not install QT or KDE packages and compile your desktop from source. It's a major pain in the ass but I did it for years. A side benefit is you can participate in testing and interact with KDE developers directly.
Another option is to go all FrankenUNIX and add neon sources to your apt cache. I've done similar but I don't recommend it.
Or you can just run unstable. Lots of people do. I did for a long time, and as long as you're willing to fix the package system occasionally it's not a bad experience. Certainly better than the two previous options.
I am more than happy to say on Plasma 6.3 instead of hopping between versions where dragging files into a browser works and next update it doesn't.
I'm too impatient to use a non-rolling release distro like Debian as my main OS, that, by my standards, already starts a new distro version with some outdated packages. I admire Debian though and it is my favorite server OS.
Testing is a rolling release. Unstable is a more exciting rolling release.
Kudos to the team.
My first contact with Linux was with Debian 2.1. Exactly with this distro CDs https://archive.org/details/linux-actual-06-2/LinuxActual_01...
To be honest, it was a miserable experience to install it on your main computer without anything else available to look for help in case of problems. It was also hard to really try it due to lack of drivers for current (at that moment) ADSL modems.
But here I am a crapload of years later, still loving it :-)
Maybe title should note that it has now been released? There has been many updates about Trixie in the past few months in preparation for today.
I think the title has been trimmed from the word “realeased”. Might be another case of HN auto title edit botching the original title.
Ooo la la, GnuCash 5.10. I'm still on 4.4.
Can't find release notes though, help?
I've found the docs, but they're huge. I'll looking specifically for a list of differences between 4.4 and 5.10. (Or at least the biggest differences)
How soon can I update my raspberry pi 5 from Bookworm to Trixie? Does PiOS have to initiate that first?
Raspberry Pi OS is a derivative and not straight up debian. It's not a released yet. A beta exists and looks like this one will support an in-place update
The images are a bit out of date, but check out https://raspi.debian.net/
Debian was often the only linux os that worked on old "spacestations" of mine. Great sentiment
And I mean Debian 12, not some old version, much more impressive
Do you mean SparcStations?
Looking forward to upgrade over the weekend.
Have had my RPi on Debian since Debian 9, with smooth upgrades every time.
Apparently Trixie needs a larger boot petition than prior versions and therefore I have to do new installs on three of my homelab machines - and probably my proxmox machine too. What a headache.
That sounds concerning, especially since Debian had hard-coded (!) a tiny /boot partition for encrypted disks for a long time. This already caused problems quite frequently (you have to manually delete them quite frequently, and which inhibits your ability to revert after a kernel regression - which, hm, I have noted to have been relatively common for Debian 12 "Bookworm" compared to usual ... hopefully Trixie is better, but if it makes the kernel-management problem harder that's a bad sign)
I have been tracking Trixie on my Resolve workstation for the past couple of months. The only hiccup was that the latest kernel did not support the ondemand governor, so I had to build a custom kernel to fix that.
I've been using the Debian trixie branch for about a year now on my local server, never once had a real issue with anything. Very impressive.
Nice, more chaos to unstable again. Sincerely happy because of the release and new stable, but also selfishly happy because now unstable starts moving again.
> The overall disk usage for trixie is 403,854,660 kB (403 GB)
That's too big. I'm going to need a smaller distro.
You're probably being sarcastic, but in the eventuality you aren't, this is if you install every single package.
No, it’s actually the disk usage of hosting a full mirror, other users commented on that
I’m sure it’s small compared to some alternatives.
You're unlikely to install everything in it.
> The overall disk usage for trixie is 403,854,660 kB (403 GB), and is made up of 1,463,291,186 lines of code.
This makes Debian Trixie about 32 times larger than Windows XP with approximately 45 millions lines of code, arguably the best Windows OS ever.
Debian Trixie is released about 24 years after Windows XP.
Sure, but XP came with minimal amounts of bundled software, that's every package in the debian repo.
> Windows XP with approximately 45 millions lines of code, arguably the best Windows OS ever.
Naah. Early(ish) Win7, or maybe even late Vista.
Congrats to the Debian team!
The Devuan version may end up being the last that GNOME will run on...
You say that like it's a bad thing.
I don't like GNOME's engineering or license, but as far as the UI and design are concerned I don't think anyone has topped it (minus the hot corner I guess).
Meh, it looks pretty in screenshots but it's a huge waste of screen space in the end.
I suppose 'power-users' can put up with 'small-button' GUIs but I do think a general purpose GUI needs to be easy to use for people who aren't used to controlling mice and keyboards all the time.
Being able to use mouse and keyboard makes you a power user in 2025?
No
Then what was your preceding comment supposed to mean?
My point is that small button GUIs should be a choice for the people who want to put up with them, not something that's pushed for universal use.
Those users won't use Devuan or Gnome but Android/iOS.
They may not personally choose Devuan or GNOME but they might end up using them because an administrator chose them.
So what is the actual difference. These release notes are not very clear. They just give version bumps. How can people get excited when you give them nothing to get excited about?
Io i
When can we have Bandit or Bluey?
Presumably not before Debian runs out of Toy Story characters.
...and since there's Toy Story 5 scheduled for 2026, the pool of yet unused characters will become larger again soon.
I can't believe we've come to such a high number, and a particularly lucky one at that
Alas it's still not suitable as a daily driver for the average home user and probably never will be. It is unfortunate that Ubuntu has to reign supreme in that regard.
Don't feed the troll, etc... But I just had to bite on this bit:
> Alas it's still not suitable as a daily driver for the average home user and probably never will be. It is unfortunate that Ubuntu has to reign supreme in that regard.
It's true that Ubuntu used to be the OOB ready version of Debian, which "just worked", while base Debian took look of fiddling to even have wifi working.
These day though I find the opposite to be true: Ubuntu does lots of weird things I don't want, and I have to "fiddle" to disable all that. A base Debian install however (ISO with firmware bundled), just works.
For me, Ubuntu is officially off my list of distros I bother spending my time on.
Two kids in 4 to 16 range, and two adults in 30 to 46 age ranges have been using Debian on daily basis for almost a decade now. At least three of them are pretty "average home user". There has been forced use of windows (since school and employers wanted), but for home use Debian has always been better due to less maintenance needs and no distractions.
You can install debian and ubuntu with same DE and you'd be hard pressed to find a difference apart from the theme unless you are a power user who knows what snap is.
In fact, Ubuntu has never been an especially user friendly distro. At the beginning it was just a debian that was installed with debian's experimental installer before they decided to use it in stable. Nothing more, nothing less.
If you wanted to find a distro that was making efforts towards beginners looking for Gui config tools, you had to look at Suse and Mandrake (now Mandriva).
The only specific thing Ubuntu did for beginners is sending CDs for free at a time when not everybody had fast internet connections and would look for paper magazine to come with CD/DVD. And they have stopped doing that a loooooong time ago.
>The only specific thing Ubuntu did for beginners is sending CDs for free
Assuming you are not malicious I will kindly help with your bad memory, Ubuntu had always very good proprietary driver support, this made laptops actually work and helped beginners. I also remember they had a graphical installer compared to Debian and for sure this was beginners friendly. Maybe some other distro offered easy way to install and come with proprietary drivers setup but I can't remember a deb based distro doing that.
Anyway you were wrong, the CDs were not the only thing made Ubuntu appeal for beginners, there were Linux magazines with CDs each month and they were not super expensive , my first linux was a Kubuntu 6.10 from a magazine and I am still running Kubuntu today though i ran Debian, Sidux, Arch, Mandriva, SUSE in the past when I had time to try different distros, compile custom kernels etc.
the graphical installer was debian's new experimental installer. They just decided to release a stable distro before debian with it.
Proprietary driver installation was the sole reason of existence of Linux Mint which was a fork of ubuntu, so your memory is incorrect.
>Proprietary driver installation was the sole reason of existence of Linux Mint which was a fork of ubuntu, so your memory is incorrect.
I think your memory is incorrect, you might be thinking of video codecs and maybe Flash not proprietary drivers, since Ubuntu already had support for easy install of drivers before Mint.
Or unless the capacity of you harddisk is limited, filling up with huge snap packages.
Or you need something that is broken by snap. I helped a user after thunderbird after an upgrade could no longer open PDF-s in okular. It turned out that the thunderbird dpkg had been replaced with a snap and I spent quite some time trying getting it to work, filed bug reports, etc, before giving up and installing from Mozilla until I replace it all with Debian.
> Alas it's still not suitable as a daily driver for the average home user
I think that's fine for Debian. Maybe even a good thing.
Debian supplies a rock solid base for many general purpose tasks. Ubuntu and other distros are free to package that up in a user friendly way, but as a technical user I want to be able to go upstream and get a basic Linux system without extra stuff.
> Alas it's still not suitable as a daily driver for the average home user and probably never will be.
Why not?
My family members need little more than a web browser, media player, and office suite. Debian Stable is very suitable here; arguably more so than other distros, which tend to require maintenance more often.
The installation is slightly easier (but still hard because of USB install) and the website has a more appealing design. Except from that what is better in Ubuntu for the average casual user? Proprietary blobs are now included in the default installer since version 12.