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
Yeah I assumed it would be misread like that. I meant it on the technical level, how equal ext4 is to NTFS in it's reliability, stability and all the technical aspects of the filesystem, knowing full well that NTFS is not meant for Linux. Basically what would be the Linux equivalent of NTFS, and if ext4 can be considered one, or does it have weaknesses that disqualify it from being considered one.
pretty much any modern linux filesystem is gonna be significantly better and more advanced than ntfs
if u ever run into a situation where u think "damn i wish i had ntfs for this particular use case" ping me cause i cant think of any
 
How worse is ext4 compared to NTFS anyways? NTFS is practically ancient by this point yet you don't hear any major problems from Windows that stem from it. If ext4 is comparable to NTFS then there isn't much reason to use anything else for your OS install IMO.

ext* has always been the safe choice, and you can even deliberately format filesystems to have just the ext2 feature set (to still be mounted by the ext4 driver on modern kernels) if you want full backwards compatibility to really old kernel versions (including those which only have ext2 support), as well as rock solid reliability.
 
Just to hammer the point home about ext4, even RHEL and its clones prefer xfs and ext4 above all other (flashier, newer) filesystems.
 
Last edited:
I'm just surprised there isn't a popularized retro exclusive Linux distribution for 32-bit and older hardware. If there really is a market for it, people should make one.
I was thinking about this the other day. Someday soon the kernel is going to drop 32bit support and a lot of old driver code has been stripped. There won't be nearly as much support for 32bit stuff so it would make sense for everyone to consolidate into a few "retro distros".

Fork the kernel (maybe even from 2010 or so), actually support i486 and up, goal be to support everything manufactured between 1995 and 2010 or so. Try not to have so much bloat so you can run a minimal install on 128mb of ram. Sidenote: it really is brutal how much early windows mogged linux in terms of usability and resource requirements.
 
Another thing that's also weird about KDE too is they have their own distro called KDE Neon. and now KDE Linux. I've never heard of anyone ever using them.
KDE Neon is their testing distro that puts the very latest version of KDE on a stable Ubuntu base. This ensures you get the latest patches as soon as they are merged.
KDE Linux is an immutable distro that intends to be locked down like Android, and to the best of my knowledge means that every update is a 2+gb download as you have to redownload the entire os every time. To be honest KDE is literally the worst DE for that system because it pushes patches so frequently that you will be constantly downloading hungreads of gigabytes every month or suffering from features being broken and out of date. Immutable works best with something like Cinnamon where updates are slow but throughly tested. I think I can see what their vision is for KDE Linux, but I don't think that's a future that will come true.
 
Yeah I assumed it would be misread like that. I meant it on the technical level, how equal ext4 is to NTFS in it's reliability, stability and all the technical aspects of the filesystem, knowing full well that NTFS is not meant for Linux. Basically what would be the Linux equivalent of NTFS, and if ext4 can be considered one, or does it have weaknesses that disqualify it from being considered one.
Ext4 is the filesystem for Linux. It's been extensively battle-tested over almost two decades and is all but guaranteed to be reliable.

NTFS is incredibly slow on Linux. Anecdotally, I've had my PCIe 3.0 M.2 drive (~3500MB/s R/W) use NTFS for historical reasons, but when I was benchmarking my CPU by compiling the kernel, on the M.2 it became bottlenecked by the I/O, meanwhile on my main drive (traditional SATA III SSD, ~500MB/s R/W) with ext4 it would work at max CPU power. Though, it could've been because I was using the FUSE driver, instead of the newer in-kernel ntfs3. For those curious: 71 seconds with defconfig, around 18 minutes with allmodconfig.
 
Last edited:
How worse is ext4 compared to NTFS anyways? NTFS is practically ancient by this point yet you don't hear any major problems from Windows that stem from it. If ext4 is comparable to NTFS then there isn't much reason to use anything else for your OS install IMO.
I believe they're roughly around the same technically/feature wise, ie. they're both journaling file systems. However ext4 is probably better at this point in time because I doubt microsoft has been putting any resources into NTFS in the past decade compared to continued ext4 support.

That said I disagre entirely with the rest of your premise. EXT4 is a legacy file system at this point. You want speed? XFS. You want fancy features + data integrity? BTRFS (the correct choice). And I have had btrfs catch multiple hardware problems over the years.
 
xfs has the benefit of being copy-on-write which means duplicate data won't write twice, saving ssd lifespan especially when using a distro like arch or gentoo which has a lot of source packages.
xfs is not a copy-on-write filesystem. It has some optional CoW features in constant development, but they're less battle-tested, and the filesystem isn't built to be CoW from the ground up (e.g., with metadata), so it will likely never be fully CoW however many new features are added.

Also, being CoW doesn't mean duplicate data will never be written twice. You can do cp --reflink when applicable, but I don't see how that could help with installing packages from source.
 
BTRFS (the correct choice). And I have had btrfs catch multiple hardware problems over the years.
Not a fan of ZFS? I wanted to do a ZFS on root Gentoo install, but it seemed not worth the effort for my use case. Apparently, from what I've been told, ZFS is only benefits from having multiple pools. Also, I read that it can cause a system freeze if you use it in conjunction with zram (at least that's what the Arch Linux wiki told me). And it can break with kernel updates because it needs an out of tree module that is very picky.
 
Glad to see LTT confirm that he is not the real Linus.

Screenshot_20250908_214222_YouTube.webp


Also, being CoW doesn't mean duplicate data will never be written twice. You can do cp --reflink when applicable, but I don't see how that could help with installing packages from source.
I have cp aliased to cp --reflink (although it's already set to --reflink=auto by default so any filesystem that supports COW will already be using reflink).

Even if it wasn't built to be a CoW filesystem originally it does a good job handling it from my use at least.
 
Where do you see this? They shouldn't interact (obviously keep zram away from your zfs pool)
Here. The Arch Linux wiki says using ZFS with swap can cause a system freeze because zvol uses memory intensive operations. I use a lot of zram as a replacement for swap because I noticed WINE games would freeze up my system consistently without any swap if I played any games for too long.
 
Not a fan of ZFS? I wanted to do a ZFS on root Gentoo install, but it seemed not worth the effort for my use case. Apparently, from what I've been told, ZFS is only benefits from having multiple pools. Also, I read that it can cause a system freeze if you use it in conjunction with zram (at least that's what the Arch Linux wiki told me). And it can break with kernel updates because it needs an out of tree module that is very picky.
I don't like ZFS simply because as you mention it's not in the kernel and that causes issues with versioning. Like, does using an out of tree module that has specific kernel version requirements to mount your root FS while running a rolling release distro sound like an attractive proposition to you?

Further I know their VDEVs need all the same sized disks (or you lose the extra space) where as my btrfs arrays I can buy whatever is the best TB/$ at the moment then chuck it in there and it just works. Want to swtich raid profiles? Single command that also just werks. (the array is accessible through all of this). In regards to the freezing I wouldn't know because I haven't used ZFS.
 
Another thing that's also weird about KDE too is they have their own distro called KDE Neon. and now KDE Linux. I've never heard of anyone ever using them.
I used Neon for a while (it came with the laptop I bought for work). It was fine, until I had to upgrade to a newer version of the underlying ubuntu base and everything broke. I switched to MX after that.
 
xfs is not a copy-on-write filesystem. It has some optional CoW features in constant development, but they're less battle-tested, and the filesystem isn't built to be CoW from the ground up (e.g., with metadata), so it will likely never be fully CoW however many new features are added.

Also, being CoW doesn't mean duplicate data will never be written twice. You can do cp --reflink when applicable, but I don't see how that could help with installing packages from source.
There's a nice benefit to using XFS on Gentoo: it supports reflinks, while ext4 doesn't, so you get nice CoW when Portage merges packages to the live filesystem.
-sam
 

Dev snared in crypto phishing net, 18 npm packages compromised​


Popular npm packages debug, chalk, and others hijacked in massive supply chain attack​

icon
Thomas Claburn
Mon 8 Sep 2025 // 19:06 UTC
Crims have added backdoors to at least 18 npm packages after developer Josh Junon inadvertently authorized a reset of the two-factor authentication protecting his npm account.

The malware targets cryptocurrency transactions on various blockchains such as Ethereum, Bitcoin, Solana, and Tron.

In posts to Bluesky and GitHub on Monday, Junon acknowledged that a phishing email had duped him, allowing miscreants to take over his account
According to ReversingLabs' 2025 Software Supply Chain Security Report, 14 of the 23 crypto-related malicious campaigns in 2024 (61 percent) involved npm, with the remainder linked to the Python Package Index (PyPI). ®
 
I've been having a two day struggle session with a new laptop. My silver lining is that I am down to a 10 minute Arch install without needing to use any reference material. I think I'm over most of the issues, but I'm curious if y'all have run into stuff like this.

On my desktop I've been running a system with a single boot partition and the rest as a LUKS/LVM setup. This worked great on my desktop when I had grub, and when I moved to systemd-boot it seemed fine with it, too. (I ran into a minor issue while setting up secure boot where sbctl couldn't find my boot folders, but that turned out to be because I hadn't flagged the partition properly). I had all my steps written down as used the notes to rebuild the setup on this laptop, and a boot would get as far as the boot loader but I couldn't even get to where it asks for the key to decrypt my LUKS partition. I'd get two messages, one stating that it failed to flush my random-seem and another stating it couldn't find the root. I found a guide that detailed the same steps I took, but broke up my boot partition into /boot (flagged as XBOOTLDR) and /efi (flagged as ESP). I still get the message regarding random-seed sometimes, but I can select and get to my kernel now.

After I got Arch set up initially I was looking at getting my Windows disk set up. I kept running into an error stating that the Windows boot loader couldn't find my /efi/ partition. If I kept restarting it would eventually catch and continue setup. I tried the install after removing the Arch NVME but still ran into the message occasionally. I'd made the installer maybe 2 weeks ago (and used it on my desktop succesfully), but figured I'd try reflashing the USB drive and made a new one. This time it worked and I was able to install Windows. Once I had it signed in I put the Arch drive back in and set up secure boot.

Once secure boot was set up I wanted to get systemd-boot set up to let me into Windows. On the desktop it was just a matter of copying the Microsoft folder out of the EFI Windows had set up and moving it into my /boot/efi/EFI/ and making a loader entry for it. Systemd-boot also found the Windows EFI on its own and added an entry, but this one was triggering the bitlocker screen, despite my having put the reboot-for-bitlocker line in the loader.conf. My entry worked though, and would actually reboot to get into Windows so it wasn't an issue. However, I've tried this on the laptop and copying the folder like I did previously causes me to hit the screen stating it can't find the drive. Systemd's auto generated entry still tries to go straight to Windows without a reboot causing it to bitch about secureboot policy changing and demanding my bitlocker key again which was the same behavior as the desktop.

So for now I'm forced to use UEFI boot select rather than just using the boot loader menu. This isn't ideal because I'd like it to default to my previous entry. Also somehow Windows got wind of the ESP on my Arch drive and wiped the whole fucking thing during a reboot last night. I had to grab the Arch installer and arch-chroot back into my install, delete the Windows entry on there and re-run bootctl install. This is... some typical fucking Microsoft bullshit, but I guess I expect this to happen occasionally. I recall bringing up Windows eating my grub install previously and there was some talk about it being a grub only thing, but I suppose not.

The thing about splitting my boot partition into an ESP and extended boot loader bothers me for some reason, but Arch is running fine so it's more just nagging at me as to why is it not working like it was on another machine. The failure to flush random-seed I believe is something to do with it trying to generate a new seed file before disks are mounted completely. And Windows bullshit. It's always Windows bullshit. Would another boot loader possibly be a better option? I'm open to it as long as it's capable of handling encrypted lvms and I can do secureboot.
 
Back
Top Bottom