The Linux Thread - The Autist's OS of Choice

  • Want to keep track of this thread?
    Accounts can bookmark posts, watch threads for updates, and jump back to where you stopped reading.
    Create account
Max Weber said:
Problems like these need to be solved at the root: libc. Drop glibc and write systemd-libcd (hand it off to Claude, it can handle it). All std calls (and especially syscalls) are checked against a database to see if 1) the code is allowed to make the call and 2) the user is allowed to make the call. If you want to change the database on your local machine to give permissions to a program ("sideload" it), you first need to register and verify your Developer Account with Freedesktop.

I hope Mr. Poettering sees this post and thinks about my proposal. We desperately need someone to finally rein in the wild, insecure, unsafe space that is Linux desktop.
Click to expand...
That is a much more elegant idea! However, it is possible to create binaries that do not use libc, musl, or other libraries of the sort, just raw syscalls. To prevent this causing a security hole, all syscalls which hit the kernel from 'unsafe' (anything which doesn't begin in systemd) programs should send a DBUS message to systemd-syscallmon to check if they are allowed. systemd-syscallmon can then make the appropriate checks and come back to the kernel to determine whether they can proceed or not. At least, that's what I'm reading in Mr. Poettring's microsoft.com email account as being his next plan.
 
⠠⠠⠅⠑⠋⠋⠁⠇⠎ ⠠⠠⠊⠎ ⠠⠠⠁ ⠠⠠⠋⠁⠛ said:
That is a much more elegant idea! However, it is possible to create binaries that do not use libc, musl, or other libraries of the sort, just raw syscalls. To prevent this causing a security hole, all syscalls which hit the kernel from 'unsafe' (anything which doesn't begin in systemd) programs should send a DBUS message to systemd-syscallmon to check if they are allowed. systemd-syscallmon can then make the appropriate checks and come back to the kernel to determine whether they can proceed or not. At least, that's what I'm reading in Mr. Poettring's microsoft.com email account as being his next plan.
Click to expand...
clearly at this point it is going to be better to implement systemd-kerneld to create fully secure systemd/systemd system
 

Depending on how Fedora turns out over the next few months, I might jump back to OpenSUSE.

The nuclear option would be Slackware, but we'll see.
 
Kuroyuri said:
The quality of OpenSUSE has been steadily declining the past few years, the people behind it are more interested in commie politics than making a solid product. :sigh:
Click to expand...
Wasn't aware. Have recently put OpenSUSE on a test system, hoping it was a stable?
 
Kuroyuri said:
The quality of OpenSUSE has been steadily declining the past few years, the people behind it are more interested in commie politics than making a solid product. :sigh:
Click to expand...
Well... shit. 😩
It's why I said "might". Fedora for now Just Works®, at least for me.
 
Kuroyuri said:
The quality of OpenSUSE has been steadily declining the past few years, the people behind it are more interested in commie politics than making a solid product. :sigh:
Click to expand...
yeah i will never understand people who glaze opensuse
i have several problems with tumbleweed which ive only been able to use briefly
1: custom kernel thats difficult to replace. i was trying to use an xbox controller over bluetooth and had to look through docs and install like 10 different user repositories only to fail to get it to connect, and it got to the point where it was looking like i was gonna need to swap kernels, but opensuse's kernel build system is not generic at all and is more hassle than its worth. i then joined a kernel development group and they had the same problem porting tkg over to opensuse.
2: yast is garbage. i know its abandonware and i found that out realizing some btrfs stuff was never implenented, and they told me oh yeah thats being replaced with something else. i really didnt want to use yast because i was used to doing things the normal fucking linux way and yast overrode some settings like the printer setup in kde, which pissed me off a bit.
3: vendors are fucking aids. i decided to try opensuse again right when they had that critical repository issue with mesa and had to wait i think over a week for the conflict to be resolved, which really made me wonder how often this would happen and why they cant just do repos like a normal distro. by the time i was done with opensuse i had like 20 user repositories installed for basic packages.
4: everything was fucking broken. mpv, broken. had to use the flatpak. github desktop, same problem. i couldnt trust the native opensuse packages would even work on my install and had to resort to flatpak for way too many things.
"stable" rolling release, my ass. its just overengineered german shit and thats all itll ever be.
 
Crossposting from OSS thread:

Ferryman said:
Holy gem. It really is a shame that the core devs of Wayland are such unabashed faggots, there's tons of good Wayland-specific software out there that would benefit from fully decoupling them, i.e. Hyprland.

Edit: peep this https://gardenhouse.pinkro.se/, its a guy making reconfigured systemd elements for standalone use. Saw it in the Guix mailing lists in reference to Gnome's ever-hardening systemd dependencies with the implication that the dev for this may be able to put together shims and such for avoiding further enshittification, both for Gentoo and Guix. Might be an interesting thing to keep an eye on as things develop.
Click to expand...
 
Fcret said:
You think the BSDs won't comply? The entire BSD community is full of far-left loony troons and infosec "experts" who would love nothing more than to have an excuse to introduce RealID to the operating system so they could ban problematic people from using it.
Click to expand...

I can see OpenBSD refusing mostly because de Raadt's a schizo. Free and Net wont even fight and are probably happy that this is happening.
 
Gulo Gulo said:
I can see OpenBSD refusing mostly because de Raadt's a schizo. Free and Net wont even fight and are probably happy that this is happening.
Click to expand...
Free and Net would take the position that, "UMM ACTUALLY IT WAS ALWAYS THE INTENTION OF THE ORIGINAL UNIX CREATORS TO HAVE A BIRTH DATE FIELD FOR ACCOUNTS BUT THE LIMITATIONS OF THE PDP-11 PREVENTED IT THEREFORE WE ARE JUST CORRECTING PAST DESIGN OVERSIGHTS"

Kuroyuri said:
The quality of OpenSUSE has been steadily declining the past few years
Click to expand...
Do not use German software.
 
Last edited:
Max Weber said:
Problems like these need to be solved at the root: libc. Drop glibc and write systemd-libcd (hand it off to Claude, it can handle it). All std calls (and especially syscalls) are checked against a database to see if 1) the code is allowed to make the call and 2) the user is allowed to make the call. If you want to change the database on your local machine to give permissions to a program ("sideload" it), you first need to register and verify your Developer Account with Freedesktop.

I hope Mr. Poettering sees this post and thinks about my proposal. We desperately need someone to finally rein in the wild, insecure, unsafe space that is Linux desktop.
Click to expand...
I have never worried much about security in my Linux install until the recent years when every soylanguage out there now has their own soypackage manager where soydevs simply soygit soypush their latest edits and the soypackage's soyversion immediately bumps up to the latest thing unverified by real engineers. I will start respecting systemd when it prevents npm, pip and cargo from running at all.

Bash: 
soypkg start-new-project
soypkg add-pkg soybean # soybean's DEV GOT PWNED 5 MINUTES AGO NOW YOU TOO ARE FUCKED

A big thanks to the JavaScript, Python and Rust community for helping normalize this
 
Unintelligence Score said:
I have never worried much about security in my Linux install until the recent years when every soylanguage out there now has their own soypackage manager where soydevs simply soygit soypush their latest edits and the soypackage's soyversion immediately bumps up to the latest thing unverified by real engineers. I will start respecting systemd when it prevents npm, pip and cargo from running at all.

Bash: 
soypkg start-new-project
soypkg add-pkg soybean # soybean's DEV GOT PWNED 5 MINUTES AGO NOW YOU TOO ARE FUCKED

A big thanks to the JavaScript, Python and Rust community for helping normalize this
Click to expand...
The infuriating other side to this coin is:

The solution to saving yourself from this fate is to become really interested in how the software on your computer is built. Where did the source code come from? Somewhere that it can be tampered with? How was the build configured?

Once this happens, you start noticing that a lot of your software isn’t built very well. This is generally for two reasons:
1) the software is being built to maximize compatibility for as many computers as possible (which isn’t a terrible goal)
2) the maintainer is lazy or ignorant

Of course, once you start suggesting improvements, suddenly you’re the asshole for thinking you know more than the maintainer. Nevermind the whole point of open-source software is to encourage freedom and software that can improve…
 
Anyone using Waydroid on Xorg/Xlibre? How well does it integrate? Can I have something like YouTube Revanced working seamlessly on my GNU+Linux(tm) system?
 
Switched to Linux ~12 years back with Ubuntu 14.04 and been using Debian as my daily driver for over a decade at this point.

I've been pretty neutral on systemd though I did notice a slowdown in boot back when Ubuntu ditched upstart for it. But them bending down to age-verification laws is not acceptable. Unsurprising since it's a corpo product from RedHat but unacceptable nonetheless.

So, what do? I use Cinnamon (which is X11) and I want software to build/run as painlessly as it does on Debian.

Options I can see are the good ol' Devuan, Void Linux and Artix. MX Linux is an option too I guess, I'll remember its maintainer for generously providing a whole repo of backported packages for Debian stable.

Artix is an interesting one because I was going to switch to a proper rolling release like Arch at some point (been using Debian Testing/Sid). But the website could be better (Debian wiki looks better honestly) and there's no guarantee it will be alive after 10 years unlike Devuan for example which has stood the test of time.
 
1774176547364.png

The youtoob algo is trying to tell me something.

:thinking:
 
Binary-Tyrant said:
MX Linux is an option too I guess, I'll remember its maintainer for generously providing a whole repo of backported packages for Debian stable.
Click to expand...
MX uses shims to make sure most systemd-infected software still works. They did temporarily stop distribution of some of their dual-init and sysvinit ISOs with MX 25, but that was reversed in 25.1 when they got init-diversity-tools working, which promises to allow easy switching between even more inits in future. It's the best debian-based distro I've personally used.
 
You must log in or register to reply here.
Back
Top Bottom