Login
You're viewing the discuss.systems public feed.
  • dataramadatarama@hachyderm.io
    Oct 17, 2026, 6:03 PM

    Something I occasionally think about, and then proceed to be horrified to my core over, is this:

    A Raspberry Pi 4 *vastly* outperforms an Amiga 4000. Or a Symbolics XL1200. Or an SGI Octane. A typical microSD card is *much* faster than a Seagate Cheetah Ultra320 SCSI HDD you might have found in a high-end Unix workstation back in the day.

    But the Pi 4 is barely able to run modern-day office software or browse the mainstream web. The microSD card is mostly considered a slow piece of junk (which it very likely is, though for other reasons than its performance).

    What the fuck are we even doing here?

    💬 23🔄 65⭐ 92

Replies

  • dataramadatarama@hachyderm.io
    Oct 17, 2026, 6:06 PM

    ("very likely", because industrial-grade microSD cards actually solve many of the problems of consumer ones. Some of them will handily outlast whatever device you put them in.

    They're terribly expensive, though - and still physically fragile, small and fiddly.)

    💬 1🔄 0⭐ 2
  • dataramadatarama@hachyderm.io
    Oct 17, 2026, 6:49 PM

    The entirety of Wirth's Oberon System - a multitasking OS, compiler, runtime and TUI library - would fit in the Pi 4's L2 cache. And there'd be room enough left over for...

    ~4 more copies of the entire Oberon System.

    💬 2🔄 14⭐ 16
  • 💬 0🔄 0⭐ 1
  • Louise 🏳️‍⚧️xorlou@lgbtqia.space
    Oct 18, 2026, 9:34 AM

    @datarama I'm still a big fan of CompactFlash, perfect form factor (not too small, fiddly and fragile) and also available as industrial grade. I can easily snap a microSD card in two with my hands.

    💬 0🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • dataramadatarama@hachyderm.io
    Oct 17, 2026, 6:12 PM

    @ireneista There was a time where I would have given a kidney* for an SGI Octane. And now, it wouldn't even be able to run boring mainstream software - despite a lot of the software it *did* run actually being terribly exciting.

    *) not necessarily my own

    💬 0🔄 0⭐ 2
  • David Chisnall (*Now with 50% more sarcasm!*)david_chisnall@infosec.exchange
    Oct 17, 2026, 6:11 PM

    @datarama

    I just set up an M1 MacBook Air for my mother. Her old laptop is starting to be unreliable. I couldn't get the Migration Assistant to work but even tying was pain. It took almost a minute to launch Edge. The installer took twenty minutes to run and install a simple smallish program.

    The machine had 'only' a 2.48 GHz superscalar out-of-order processor with over 2 MiB of cache and 4 GiB of RAM.

    The RAM was over 90% used immediately after boot and the CPU was sitting at 100% with no user-visible applications running. The machine, after boot, had 300 processes running with an average of ten threads each.

    No one who was involved in shipping any if that has any business referring to themselves as anything engineer-adjacent.

    💬 0🔄 0⭐ 4
  • Book Squirrelbooksquirrel@tech.lgbt
    Oct 17, 2026, 6:16 PM

    @datarama Just like traffic volume expands to fill out available roads, so does software size and complexity increase to fill out available resources, regardless of any added utility I'm pretty sure.
    I'm not entirely convinced that we have gained much of actual use from the past some 20 years of chip development (apart from a number of highly specialized use cases).

    💬 1🔄 0⭐ 2
  • Book Squirrelbooksquirrel@tech.lgbt
    Oct 17, 2026, 6:47 PM

    @datarama Just to illustrate: Depending on what you want to do with Blender, it can benefit from some fairly beefy hardware. But!...
    I went to university from 2004 to 2009. As you may remember, this was a time during which I was having a lot of fun learning to use Blender for the purpose of creating some weirdly overengineered movie posters for the movie club.
    So I could run some actually quite advanced 3D modeling software back then, on hardware affordable to a student.
    Even at that time I feel we were moving (or had moved already) well past the kind of hardware most of us would ever actually need.

    💬 0🔄 0⭐ 1
  • 💬 0🔄 0⭐ 0
  • 💬 1🔄 0⭐ 1
  • 💬 0🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 1
  • 💬 0🔄 0⭐ 0
  • Stutehstu@hachyderm.io
    Oct 17, 2026, 6:48 PM

    @datarama I've "dailied" my Pi 4, and then later Pi 5*, since launch and I agree. Admittedly, the 5 eeks out just enough performance that it becomes acceptable with modern software if you're patient, but it's still a completely ridiculous situation.

    *with Ubuntu

    💬 0🔄 0⭐ 1
  • rk: it’s hyphen-minus actuallyrk@well.com
    Oct 17, 2026, 6:50 PM

    @datarama

    I’ve often wondered how difficult it would be to build an Oberon system with a more traditional WIMP interface. I love Oberon and the Oberon system but on a touchpad the mouse chording is terrible.

    💬 0🔄 0⭐ 0
  • marocddotmarocddot@hachyderm.io
    Oct 17, 2026, 6:51 PM

    @datarama Oberon was a gem for doing stuff that the mainstream didn't think was possible and didn't think they wanted.

    Wirth and Chuck Moore were GOATs at demonstrating how bloated most of the mainstream was.

    💬 0🔄 0⭐ 2
  • 💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 1
  • rk: it’s hyphen-minus actuallyrk@well.com
    Oct 17, 2026, 6:57 PM

    @millihertz @datarama

    In a minor bit of defense for the newer software: accessibility and complex text layout for non-Latin writing systems is computationally expensive and didn’t get good until…jeez 2010 maybe.

    But we dragged a lot of other bullshit in on top of it.

    💬 2🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • SOMETHING WONDERFUL HAS HAPPENED TO YOUR leah :oh_no:millihertz@oldbytes.space
    Oct 17, 2026, 7:01 PM

    @rk @datarama seriously, i wonder whether the whole approach to text rendering is wrong, and we should really be doing it by word, with each word taking its own approach to layout - and even its own language, input key combination and library where necessary

    of course, there's bound to be at least one language where that doesn't work out, but as far as i'm aware (which really isn't very far at all, so sorry), way more written languages have a concept of distinct words than have a concept of small, discrete alphabets making up words

    💬 1🔄 0⭐ 1
  • David JONESdrj@typo.social
    Oct 17, 2026, 10:09 PM

    @millihertz @rk hmm, i'm not sure, but. A state machine is useful (for attaching marks, and modifying the shapes of glyphs in arabic, and potential in Hangeul (although i don't think anyone does it like that)). And if you _start_ from that position, it could be somewhat simpler than the current mess. The OpenType Layout "state machine" is not good enough (can't swap glyphs, basically), but apparently the Apple Advanced Typography one is. But it's been deprecated for ages. But it suggests a path to potentially better territory.

    💬 0🔄 0⭐ 0
  • dataramadatarama@hachyderm.io
    Oct 17, 2026, 7:03 PM

    @rk @millihertz There is that. If anyone had taken internationalization seriously back then, we'd probably have ended up with dedicated UPUs. (Unicode Processing Units).

    (ISO8859 was simpler, sure, but I would go so far as to say it was a non-solution. It's hardly elegant to not be able to correctly spell a Danish and a Polish name in the same text.)

    💬 1🔄 0⭐ 2
  • 💬 1🔄 0⭐ 1
  • rk: it’s hyphen-minus actuallyrk@well.com
    Oct 17, 2026, 7:11 PM

    @millihertz @datarama

    As someone who has written more text editors than the average bear we keep recreating this problem. A Unicode code point isn’t one character because you have combining codepoints, whatever abomination is going on with emoji, etc.

    You could do something obscene where you use int32’s to store precomposed units and leak into the PUA/>0x10FFFF range for precomposed emoji or whatever but good god what are we doing.

    💬 1🔄 1⭐ 0
  • SOMETHING WONDERFUL HAS HAPPENED TO YOUR leah :oh_no:millihertz@oldbytes.space
    Oct 17, 2026, 7:27 PM

    @rk @datarama i think there's a need to distinguish between computation formats and serialisation formats

    for serialisation, Unicode is probably the best compromise we have. it still sucks, but not as much as anything else.

    for computation, then as you say, an internal index of precomputed units, mapped to and from Unicode by the OS - all userspace sees is a string of units, each of which is guaranteed to be unique and distinct

    for things that have to deal with Unicode code points, as opposed to linguistic units, uint32 leaves a flat 21-bit (thank you, Unicode, for declaring this fixed for all time!) representation. and 11 spare bits to indicate exactly how this letter is connected to the letters ahead of and behind it (such that a letter with all of those 11 bits clear is defined as the end of a combined set) according to the rules pertaining to that code point. but everything is very strongly encouraged to only deal in words (send a word to the text renderer, get a rectangle with that word in it and a padding margin for whitespace back)

    if every language has its own rendering library, then emoji can be treated as just another language, which at least keeps the explosion contained

    💬 1🔄 0⭐ 1
  • rk: it’s hyphen-minus actuallyrk@well.com
    Oct 17, 2026, 7:30 PM

    @millihertz @datarama

    Ya know. We could do the UTF-8 trick. The number of initial one bits says how many following code points are part of this unit. I don’t think we’d ever really need all 11 bits.

    Still requires an O(n) walk though. I was thinking a preprocessing stage where you gather all the composed characters and give them unique IDs in the extra space.

    💬 0🔄 0⭐ 0
  • Xerz 💗xerz@masfloss.net
    Oct 17, 2026, 7:00 PM

    @datarama I learned to enjoy it. Let them assume supercomputer-like resources are essential (which tbh for some tasks like 3D or static analysis they are, so) in order to keep prices consistently low, then keep building efficient stuff

    right now the main issue I see is preserving access to knowledge on the web, otherwise we'll be alright

    💬 1🔄 0⭐ 0
  • Xerz 💗xerz@masfloss.net
    Oct 17, 2026, 7:01 PM

    @datarama and yeah, we're now at the point where we hit our first constraint, which is RAM

    I think we'll see a lot of progress on that aspect pretty soon, hopefully enough of us accumulated enough hardware by now

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 1
  • Donal Fellowsdkf@cyberplace.social
    Oct 17, 2026, 8:11 PM

    @datarama What are we doing? Chrome.

    People are literally refusing to code things actually efficiently, and are instead claiming that doing everything in React and Electron (i.e., Chrome with some outermost browser bits filed off) is Good Enough. BUT IS IT? I CLAIM NOT!

    Coders are very good at not considering the costs of their decisions. Like at all. Even compiler developers are like this (witness LLVM).

    💬 0🔄 0⭐ 0
  • Mad Engineeringmadengineering@hachyderm.io
    Oct 17, 2026, 8:13 PM

    @datarama You used to have to spend a lot of expensive (in both time and money) programmer time doing optimization so that your programs would actually run on the average user's computer, but when RAM got cheaper, too many developers decided they weren't going to bother with that.

    It's sloth, and maybe a little greed, just like most pollution in the modern world.

    💬 0🔄 0⭐ 0
  • Tijgertje1987Tijgertje1987@mastodon.online
    Oct 17, 2026, 8:22 PM

    @datarama
    Unfortunately efficiency is not being rewarded :(
    I learned to program on a P2 or P3 (can’t remember) and always try to get my programs efficient within the constraints.

    Sure I use ram cache to prevent I/O but being efficient is just important

    💬 0🔄 0⭐ 0
  • HabrokHabrok42@chaos.social
    Oct 17, 2026, 9:41 PM

    @datarama Another comparison to add to your examples: the performance of a Raspberry Pi 2 is about the same as the performance of a Cray-2, but nobody would think of using a Raspberry for the tasks Crays were used in the 80ies.

    💬 0🔄 0⭐ 0
  • benbenjamineskola@hachyderm.io
    Oct 17, 2026, 10:29 PM

    @datarama Yeah I was thinking about something similar recently. I ran Gentoo for a bit 20+ years ago on hardware that wasn’t really high-end at the time, and it took a while to compile everything but it was definitely manageable: hours, not days/weeks, to get a functioning desktop.

    Recently using it again, on hardware that’s a couple of years old, and certain packages still take hours to compile. I’ve got like 64x the RAM and 8x the processors compared to 2004; it’s not reasonable for a browser to take four hours to compile.

    Like, you mention office software — which feels like a perfect example of something that hasn’t fundamentally changed in decades. Does LibreOffice do anything now that it (or its predecessors) didn’t back then? Maybe a little; but we’re using hardware that’s not just a little more powerful but many times more. It’s incomprehensible to me.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • slashagreeable_landfall@mastodon.social
    Oct 18, 2026, 3:51 AM

    @datarama It's all the middleware. Consider that Windows XP's install CD only used about 200MB of the 600-700 MB capacity. (And I'm *still* wondering why I had to use a 3.5" floppy to add drivers...surely they had room for that?)

    Today, 600MB *updates* are common.

    I wouldn't mind so much if all those bytes were used for security, privacy, or accuracy. But they're used for kludge on top of kludge, 4 layers of GUI on top of Win 7, and surveillance for ads.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Louise 🏳️‍⚧️xorlou@lgbtqia.space
    Oct 18, 2026, 9:39 AM

    @datarama Most of what people do today could be done on something like a 200 MHz Pentium II and 128 MB of RAM, if the software was written well. Modern hardware is massive overkill but writing software carefully and thoughtfully is out of fashion.

    💬 0🔄 0⭐ 0