Skip Navigation

How Long Should Hardware/Software Support Last?

A while back there was some debate about the Linux kernel dropping support for some very old GPUs. (I can't remember the exact models, but they were roughly from the late 90's)

It spurred a lot of discussion on how many years of hardware support is reasonable to expect.

I would like to hear y'alls views on this. What do you think is reasonable?

The fact that some people were mad that their 25 year old GPU wouldn't be officially supported by the latest Linux kernel seemed pretty silly to me. At that point, the machine is a vintage piece of tech history. Valuable in its own right, and very cool to keep alive, but I don't think it's unreasonable for the devs to drop it after two and a half decades.

I think for me, a 10 year minimum seems reasonable.

And obviously, much of this work is for little to no pay, so love and gratitude to all the devs that help keep this incredible community and ecosystem alive!

And don't forget to Pay for your free software!!!

50 comments
  • What do you think is reasonable?

    As long as possible unless nobody uses it for cases that need any security (daily driver, server, enterprise etc). If you drop support, you are lazy and support ewaste creation. In some cases it can be too difficult to support it but "too difficult" has a lot of meanings most of which are wrong.

    I think for me, a 10 year minimum seems reasonable.

    That's really not enough. GTX 1080 is an almost 10 years old card but it's still very competitive. Most of my friends even use 750s or similar age hardware. And for software, any major updates just make it more enshittificated now lol.

    • In principal I don’t disagree.

      Problem is supporting everything requires work and effort which isn’t funded by a corporation or anything

      • perhaps we should start building things with long term support in mind, and not just churn out the cheapest shit we can manage.

        Like just look at modern laptops, most of them are absolute dogshit in terms of repairability and then you have the framework which you can straight up buy as a kit to assemble yourself.
        Making things easy to maintain is clearly doable, not even that hard.

      • Hardware support is usually funded enough or has enough human resources for it not to be a big problem imo. It's ok to drop 30 years old stuff that nobody uses but dropping something just because rich people have a few years newer hardware is bad.

    • I think it should be supported for a decade and the open sourced so that it can be archived and maintained by those who care.

  • I feel like with libre/open source software, this is a lot less of a problem -- So long as it is still possible to add it back by messing around under the hood, we are pretty much fine with the "Main" branch of some software dropping legacy support?

    It'd be unreasonable to expect the devs of anything to keep supporting things that are over 20 years old.

    And like, if you're using 25 year old kit at this point you're either a hobbyist collector of vintage stuff, OR an enterprise with mission-critical assets on old legacy hardware/software -- In either of those scenarios, "figure out how to go under the hood and fix stuff" (or in the enterprise's case, "hire someone who does that for you") is not an unreasonable expectation to have.

    The smelly part is of course proprietary software and hardware, where "dropped official support" might as well be the signing of a death order. We desperately need a "right to repair and maintenance" regulation on every country in the world.

    • I'll add that at this point, if you're a hobbyist collector of vintage computer hardware, and you find satisfaction at making that old Compy 386 run like it's modern hardware, you should know how to compile your own kernel.

      Like, it just seems prudent, given the fact that it's unreasonable to expect a "universal" kernel to simply grow and never prune anything (which I think avoiding having a giant kernel was part of the rationale, iirc), and there's plenty of documentation out there on how to do it. If you aren't going to run the same hardware as 95% of your peers, it's your responsibility to make sure your hardware works.

      • Yeah I mean

        Hobbyist collectors of typewriters (I know because my father is one) and cars (one of my friends is one) all have to learn how to maintain and service their own stuff because businesses that did that for them have all but disappeared. It's considered part and parcel of the hobby.

        It'd be nuts to expect it to be any different for computer collectors. Compile your own kernels, diagnose your own problems, fix your own shit. That's what you do for a hobby. :P

        If you're running something that old, then it is by choice anyway, hardware gets more expensive after a certain age, and you definitely won't be getting a (functional) 90s computer for cheap.

  • i use 10 year old hardware and its pretty capable on linux

    we reached a point of diminishing returns in the advance of this technology

  • The thing is, Linux always gets touted as the way to save old hardware. Win 11 not supporting a bunch of perfectly good older computers is leading to a massive e-waste wave. I understand that kernel devs mostly do it for free, and resources are limited for maintaining support for hardware few use anymore, but I think having a way to viably daily drive old hardware is really important for reducing e-waste and also just saving people's money. I don't like buying new tech unless it's to replace something beyond repair—ie not just an upgrade for the sake of upgrading.

    Obviously the problem is more socially systemic than just the decisions of Linux devs. I think the release cycle of new hardware is way too quick—if it were slower obviously that would reduce the workload for kernel devs, so hardware could be supported for longer (as they have less new hardware to work on supporting). And more generally we should have a mode of production not centred around profit, so that people don't get punished (as in, they're not getting paid but could be compensated for their time if they worked on something else) for spending time developing kernel support for old hardware.

  • support should drop when we hit the sweat spot where the energy saving from running modern RISC devices over old CISC outweighs the energy cost of manufacturing replacements.

50 comments