darry wrote on 2025-05-13, 01:22:

The point I was trying to make is that, while using Linux can keep old hardware usable and useful, there comes a point where eve […]
Show full quote

The point I was trying to make is that, while using Linux can keep old hardware usable and useful, there comes a point where even pedestrian consumer appliance type hardware like a somewhat dated Chromebox can

- be faster
- run cooler
- be cheaper to operate

than very old high end desktop hardware, depending in the actual use case.

I think the future of linux, and OSes for 'desktop' generally is 64bit PCs and various SBC or similar low powered devices with an increasing trend towards everything being "cloud" and "AI" with the devices being always online, being basically high spec input/output devices. That's a guesstimate - difficult to strongly predict anything.

https://www.techradar.com/pro/pc-sales-are-ri … leading-the-way

PC sales were up again recently.... You can find market analysts that consider the volumes to stay up long term too, whether that means Linux can keep up and whether the concept of having a PC offline most of the time (like we did back in the day) will seem strange i dont know

Just noticed this video on YT:
Linux Kernel Rapidly Removing Hardware Support Because of AI / LLM
https://www.youtube.com/watch?v=8vDSSxaQ7ZU

This resonates with me, the idea that Linux would save old and obscure hardware was what I respected most about Linux.

That guy is basically mentioning certaing things that I mentioned earlier.

32bit still has many uses however doesn't have the interest it once had and isn't being manufactured much any more.
The only 32bit CPU’s I’ve seen for many many years are only for specific applications outside personal computing.
Perhaps it people need or want a 32bit kernel start your own community Linux Distro to keep a kernel fork for it.
I'm waiting for 128bit CPU’s tbh.

Yeah, been watching and tracking all the shenanigans going on with the kernel of late and removal of legacy drivers that have been STABLE for years is a bit worrying.

Removal because AI likes to poke at things it shouldn't and is fascinated by the older driver code.

Might be time to start making archives of the older distros that have full legacy hardware support.

Trashbytes wrote on Yesterday, 08:10:

Yeah, been watching and tracking all the shenanigans going on with the kernel of late and removal of legacy drivers that have been STABLE for years is a bit worrying.

Removal because AI likes to poke at things it shouldn't and is fascinated by the older driver code.

Might be time to start making archives of the older distros that have full legacy hardware support.

Older linux kernels and distros are always preserved and it's not like you are actually benefiting from anything in a modern kernel when using an i486 so I really fail to understand what the big issue is. I never understood the big hardon linux nerds got from installing Arch on a 486. Like, good fucking job, what are you going to do with it, watch Youtube? Just install a decent contemporary kernel and enjoy.

Jo22 wrote on Yesterday, 05:48:

Just noticed this video on YT: Linux Kernel Rapidly Removing Hardware Support Because of AI / LLM https://www.youtube.com/watch? […]
Show full quote

Just noticed this video on YT:
Linux Kernel Rapidly Removing Hardware Support Because of AI / LLM
https://www.youtube.com/watch?v=8vDSSxaQ7ZU

This resonates with me, the idea that Linux would save old and obscure hardware was what I respected most about Linux.

That guy is basically mentioning certaing things that I mentioned earlier.

Thankfully, there is always NetBSD, which has portability across the widest and most obscure platforms as its main goal.

And to be fair, there are some older platforms and machines out there that may have been technically supported by the Linux kernel, but were dropped in most (if not all) common distros years ago. I think this is particularly the case for big-endian architectures such as SPARC and PowerPC.

For instance, there is a Sun Blade 150 workstation at my Dad's house that I've been wanting to dust off, upgrade as much as possible and breathe some new life into. I have already decided to go with NetBSD for that machine, since I want to try installing more modern hardware in its PCI slots, and only NetBSD has strict cross-platform portability (also in its hardware abstraction layer and in all of its drivers) in its design. I already assumine that I'd be running in to way too many issues getting Linux, Solaris or Illumos to work on it.

Although I might try various operating systems and distros for comparison, once I at least have it working with NetBSD.

Jo22 wrote on Yesterday, 05:48:

This resonates with me, the idea that Linux would save old and obscure hardware was what I respected most about Linux.

Last time I checked, modern Linux was useless for GPU acceleration on hardware below OpenGL 2.0. So not really.

Jo22 wrote on Yesterday, 05:48:

Just noticed this video on YT: Linux Kernel Rapidly Removing Hardware Support Because of AI / LLM https://www.youtube.com/watch? […]
Show full quote

Just noticed this video on YT:
Linux Kernel Rapidly Removing Hardware Support Because of AI / LLM
https://www.youtube.com/watch?v=8vDSSxaQ7ZU

This resonates with me, the idea that Linux would save old and obscure hardware was what I respected most about Linux.

That guy is basically mentioning certaing things that I mentioned earlier.

Trashbytes wrote on Yesterday, 08:10:

Yeah, been watching and tracking all the shenanigans going on with the kernel of late and removal of legacy drivers that have been STABLE for years is a bit worrying.

Removal because AI likes to poke at things it shouldn't and is fascinated by the older driver code.

Might be time to start making archives of the older distros that have full legacy hardware support.

Or, here's a thought... They could just stop using LLMs for everything and use their actual braincells.

I know good 'ol Linus has his incredibly stupid hot takes but he can't really defend "generative AI" / "large language models" as a "tool" when the very way those things are engineered they are deliberately by design; false.

He and anybody else in the "upper echelons of power" could have just put their collective foot down and prohibited anything LLM related whatsoever regardless of open disclosure, since it seems to me it's been determined it's causing a problem and causing people's collective intelligence in various matters to drop off a cliff.

I certainly remember this older thread and the news that i486 was deprecated and the technical reasons why. This did not come too much as a shock, although I have been curious why they didn't just accept some minor maintenance on the core kernel.

Very recently, I remember a discussion on how AI was affecting the kernel maintenance and how it was leading to some concerns that devs wanted to drop some code. Now seeing the latest PR, they want to just rip out all these old drivers, basically giving up.

It's pretty clear, AI has made us (some people) stupider.

Going back through this thread on when it was written, it was pretty certain that the old industrial and server use of Linux (which Linux has been big in from the very beginning) had likely moved on from i486, leaving almost no one left out. This latest move on drivers likely affects a whole range of Linux users -- industrial, server, and maybe even desktop and vintage users, still breathing life into old stuff for various reasons. This is going to affect many people. So this move is certainly against the long tradition of Linux. There was no reason to do this, unlike the slightly poorly justified i486 support being dropped. They just gave it up because it is "old".

I expect a bunch of this stuff is going to start hitting the market from server and industrial applications where linux has a hold in not too long, at a really bad time due to the supply crunch.

Yeah I don't care about the i486 support but these drivers they want to remove are 100% stable and are used to provide legacy support to more than just CPUs, they range from keyboard, mice, displays, chipsets etc

These are core fall back drivers for when you have no other option for support.

Could be a nothing burger but could also cause a good number of headaches for older systems that need to remain patched and running but don't need newer hardware or don't have UEFI hardware support. Older production systems comes to mind here, have a ton of them where I work where they do one specific task and they do it so well they simply don't need upgrading to newer hardware and cant be upgraded as the software itself expects hardware that wont work on newer systems.

These systems still need kernel updates and patching for security.

Many distros are sunsetting 32 bit support entirely, a 32 bit legacy fork would make sense compared to piecemeal trimming of older 32 bit CPU and logically 32 bit only needs LTS security updates as it won't be getting new hardware

Kernel 7.0 removed VLB support from the pata_legacy driver, specifically ht6560a/ht6560b/opti82c611a/opti82c465mv/promise 20230c/20630/qdi65x0/winbond83759A.

Kernel 7.1 removed the ability to build a kernel for the 486. Crucially, at time of writing, the code needed to support the 486 (to cope with its missing CMPXCHG8B and RDTSC instructions, and FPU emulation for the 486SX) is still in the mainline kernel and the earliest it can be queued up for removal is kernel 7.3.

That means it's currently trivial to restore 486 support and someone has stepped up to maintain such a fork. Otherwise the latest maintained kernel with 486 support would be 6.18 LTS. If the pattern follows previous LTS kernels, its EoL is likely to be the end of 2030. CIP (Civil Infrastructure Platform) are unlikely to maintain this kernel after that (since they pick every other LTS kernel) but they will be maintaining 6.12-cip until mid 2035.

Systemd/linux has become a total shitshow over the last decade.

AI slop and Rust garbage in the kernel combined with a cancerous abomination of a userland.

I am strongly considering migrating to OpenBSD. Only trouble is the block layer, will have to keep one machine on lunix to run my raid6.

I said it earlier in this thread and will repeat, systemd is not an integral part of linux. There are many distros that still use initd, just use those.

But why fight City Hall and run an obscure systemd-less, glibc-less, runs-in-32MB, 32-bit Linux distribution, when you can just run stock NetBSD (albeit you'll probably need a custom kernel)?

jakethompson1 wrote on Today, 07:15:

But why fight City Hall and run an obscure systemd-less, glibc-less, runs-in-32MB, 32-bit Linux distribution, when you can just run stock NetBSD (albeit you'll probably need a custom kernel)?

Because BSD is not linux. I appreciate BSD and love the fact that it exists, but never used it and do not have the time or energy to learn it at my age..