Collapse OS
collapseos.org
collapseos.org
Also, I coudn't get this Forth to do float or double operations.
On the term size, edit cos-grid.c as en exaple for cwm:
#define WCOLS 64
#define WLINES 16It's quite plausible by looking at 2 different trends, one medium-term, one longer.
Silicon fabbing is reaching the end of the line. Not because it's no longer possible to go smaller; it is. Because of how extremely expensive it is to go smaller.
Moore's Law is not about technology. It's about money. It can be paraphrased as "the cost of putting a given № of transistors on a unit area of silicon halves every 18 months."
That stopped working linearly some time around about 2006-2007. Not because of technology; because of usefulness. Around then adding more transistors to a CPU stopped making it much faster. So instead, around then, we got multiple cores.
But that didn't pay for long, because 4 cores is about all the average desktop computer can use. Thus, SMT and hyperthreading: lie, tell the consumer they have 2× as many cores, because that is much cheaper than giving them 8 cores that they can't really use.
Server chips get more, sure, but that stops scaling quite quickly because of memory bandwidth etc.
Why does this matter? Because the silicon industry needs to pay vastly more money for smaller & smaller process tech, and for that to keep going, it needs to sell more and more chips, and if they aren't much faster, it's not worth it.
(I type on a 2016 laptop that isn't much quicker than my ~2010 laptops. My 2021 laptop is quite a bit faster, but it has a horrible keyboard and a horrible pointing device and it doesn't work well with Linux.)
Summary: around when PC chips went 64-bit bit was the last big leap. After that, more features on CPU cores didn't repay with more performance, so we switched to multiple CPU cores. That stopped paying too; now it's marginal on servers only.
CPU performance stopped ramping so fast, and the computer industry filled the gap with more RAM and SSDs. But now those don't buy that much any more either. GPUs helped but only for gaming, cryptocurrency mining (which can DIAF ASAP, TYVM), supercomputers and other niche stuff.
Problem.
It's more expensive to build modern CPUs than ever before and soon there'll be no profit in it.
Because the computer industry in general has about as much ability to plan ahead as a horny locust on meth, the result of that will be bankruptcies, massive layoffs, collapse of industrial sectors and so on.
But computers are already being overtaken by phones.
Apple is riding that wave, using very-highly-integrated phone SoCs in desktops and laptops. You can't change the GPU, you can't add RAM, you can't even easily add more SSD, but it doesn't matter (for now) because they're very fast and run cool and batteries last a long time.
And it can surf that wave for a while as all the apps switch over to Arm-native.
The PC industry can catch up: one, as other Arm vendors catch up, and then as the OSes and apps go Arm-native...
But then it's over. No more open GPU market, no more upgradable PCs.
And this very-high-level integration only buys a bit more time. It doesn't fix the spiralling process costs and so on.
The writing is on the wall. Mene mene tekel upharsin.
Issue № 2, longer scale.
There are only a handful of companies that can fab modern chips, and a smaller number who can make the chip-fabbing machines.
As climate change spirals out of control, that gets worse fast. If Taiwan becomes uninhabitable, there is an epic global problem. Look how some floods in Thailand disrupted the world HDD market for a year or so in 2011 and helped the shift to SSDs instead.
Chip fabbing is a very niche, very expensive industry. It is super-fragile. Disrupt the status quo badly and it will be one of the first to fail.
That's why the Russian army is looting washing machines: for controller chips. Russia can't make them and now it can't buy them.
It won't take a lot for there to be no more modern chips, at all.
2020s software is epically bloated and complicated. It only scales at all because of throwing vast resources at it.
If we had to go back to 32-bit chips, we'd be very screwed indeed. But when the West Antarctic Ice Sheet goes -- not "if", when -- that will cause vast global disruption and ~5m of sea level rise, globally, inside 2 weeks.
No more new silicon. At all, for decades.
Nobody can bootstrap fabbing 32-bit chips from scratch now. The supply chain needed is global.
We'll be back to hand-routed 16-bit chips at best, with a max 4GB of RAM, and not a lot of storage.
We will have to reboot the entire computer industry, and start over, because modern FOSS can't survive without fast 64-bit machines in the millions, vast redundant farms of them.
How wrong you are :)
A distro like Hyperbola runs perfectly fine on 32 bit machines with modernish browsers and GL 2.1 -at least-, enough for a lot of stuff. You have lots of lightweight GTK2/3 and QT5 software which can run fast enough even on a GB machine and less than 16GB of storage.
We don't need to reboot nil, but stop using Electron crap and JS for everything.
Half of the Gen-Z ers do today with 8GB of RAM at least to run Slack and such we, the old farts, could do that with 512MB of RAM and Kopete, with chats, videoconference, inline videos and LaTeX rendering the equations without even sweating. And Kopete IM and KDE3 where the bloated environments. We should stop wasting resources.
But the majority of them are very niche, quite obscure, and often don't live long, or splinter and factionate.
I know that it doesn't _have_ to be like this.
But Linux is most definitely not a model of lean mean software design, and it _can't_ be.
Among my favourite OSes that I've used as my primary platform at various points in the past are...
• Acorn RISC OS (fast, productive, GUI multitasking on an 8MHz ARM3 with 1MB of RAM and a 20MB hard disk)
• Psion EPOC16 (rock solid stable multitasking keyboard-driven GUI OS on an 8086 wirth 128 kB of total system storage in early machines)
• Psion EPOC32, AKA Symbian (again, very solid Internet-capable touchscreen GUI on a 16MHz ARM with 4MB of total storage
• OS/2 2.0 (pre-emptive multitasking PC OS, usable on a 25MHz 386SX in 4MB of RAM)
• BeOS 5 (gorgeous rich media OS, fast and stable on kit like a 133MHz Pentium in 128MB)
In the UNIX family, I'd point at Plan 9, Inferno, and Minix 3 as OSes far lighter and yet richer and more capable than Linux.