The computer built to last 50 years

Certainly a C64 is highly restrictive compared to a modern machine and were one to specifically build a computer to last 50 years it's not where you'd start but surely a machine that has actually lasted almost 50 years and remains usable has things to teach you about long lasting computer design.

In particular interesting to see how open source fits in. The modern C64 ecosystem has plenty of tools and utilities that do use open source software and hardware (e.g. the Kung Fu Flash cartridge: https://github.com/KimJorgensen/KungFuFlash) but plenty of the core software, that actually runs on the machine, is proprietary software the source is long gone for. It's still around because of archivists and pirates and can continue to be used because the original copyright holders don't care to enforce their copyrights. So is open source actually a core item as the author asserts or just a nice to have? Having the software be archived and easily available later was the key. Along with simplicity, you just run the monolithic binary, there's no dependencies and the software is sufficiently simple that hacking around with the raw binary is perfectly feasible.

The reason computers slow down and stop working worth a damn has nothing to do with the hardware, it's the operating system's receiving "updates" that make them quit working. I have a TRS-80 Color Computer running the Microsoft BASIC "operating system" that Bill Gates wrote himself and it still works great 40 years later.

And then the big issue with phones are the batteries. The phone manufacturers know that the batteries go bad, so they glue them into the phones so you can't replace them. Obviously if you wanted the phones to last a long time, you'd make it to where you can put a new battery in the damn thing. They also know that the screens break, so they'd make those easy to replace yourself as well if they cared.

That is nice you can take phones to those little repair places and they seem to do a nice job replacing screens and batteries, but they could probably design a phone where you can do it yourself.

https://c3.ndc.nasa.gov/dashlink/static/media/other/Observed...

The long story short is that there are byzantine failure methods which prevent a 50 year computer. A sample:

- Capacitors can act as bullets

- Forced air cooling creating water

- The smaller the parts, the greater the chance they'll transmute to another part. Even, or especially in solid state parts.

- Digital isn't (i.e. 1 isn't really full voltage, and 0 isn't really no voltage).

- Thermal expansion matters, even for ICs on a board.

- Wire length, and the position of sensors on that wire, matters.

A 50 year computer would probably have to be one in which each part can and is replaced on a schedule. And the faster the computer is, the more often parts would need to be replaced. Additionally, if we want 100% uptime there would also have to be sufficient redundancies to ensure that the computer could continue operating during failures or replacements of components.

You can have a computer with 10 years that can run modern OSs and software without being incompatible or too slow, a thing totally impossible 20 years ago.

If you do AI related developement or play videogames, you would require at least a new GPU, but outside that, I think the only couple things (pretty major IMO) making those computers less useful are more complex video formats not available to decode by hardware, and the vast amount of code some web apps use (try using YouTube or Twitter in an old laptop)

Nevertheless, it is quite easy to be able to use a modern computer for 50 years, if you just get 10 computers that do not contain components that age even when they are not used, e.g. batteries or electrolytic capacitors, and you use one computer until it breaks, keeping the others in storage until you must replace the current work computer.

Such a set of modern computers would be faster, cheaper and smaller than a single computer in the style of PDP-11 or VAX, made by using low-density components that can work for 50 years.

So arguably we've already built computers that last 40 years. Another decade doesn't seem crazy.

At first blush, this sentiment appears to also be true of old computers. There is growing "trendy" interest in them, and they're otherwise still fit for purpose for some tasks, like gaming, writing, driving long-unsupported hardware or software. The community around it has been rather industrious in servicing old machines, particularly Macs.

But they cannot satisfy all the requirements we have of a modern computer, and neither can a typewriter. However, the length of time a computer has before being truly obsolete seems much longer now than it used to be. You could easily get a decade or more if you can control the itch for new and shiny and have modest performance needs.

Might need to replace the battery, if the device has one. There's some luck involved with getting the longest support window possible from MS or Apple. Google and co are famously a lot worse on this front, if we're talking phones.

As it is now, the main motivation is "save your attention, your wallet, your creativity, your soul and the planet" which to me sounds like no specific purpose, which makes it hard to imagine if the result will have any use in 50 years.

As a comparison point this story about warning signs [0] makes the challenges a lot more palatable, and that's how I'd see any chance of success for a product design.

And that makes me wonder how many actual long lasting computer projects already exist in the world, for instance to control nuclear reactors, to activate water pumping stations, control emergency valves etc.

[0] https://99percentinvisible.org/article/beyond-biohazard-dang...

I was quite impressed to learn about the 66 year-old computer that is still in use with the Japanese transit system.

Add an ethernet card and perhaps a Zorro III RAM card, and it's usable even on the modern Internet: modern TLS works, and for sites that are too complex for AmigaDOS browsers, there are public proxies that can help.

While I wouldn't suggest anyone tries to get serious work done on the modern Internet using an original, unaccelerated Amiga 3000, it makes an excellent example of how things really haven't changed aside from speed and size since we moved to 32 bit CPUs with MMUs.

Something like this could easily be used for non-Internet heavy tasks for fifty years. We just need to be aware of the things that typically fail, such as bearings and capacitors.

[0]https://en.wikipedia.org/wiki/AlphaSmart

Offpunk. Slrn with slrnpull and mutt +mbsync/msmtp. Heaven.

Offpunk:

https://sr.ht/~lioploum/offpunk/

The point is that having the reliability of older computers might require the older, simpler hardware they were using. More breakage might be an inherent tradeoff of using modern processes, at least up to a point.

I still have always wanted to see someone mix low-cost components with the older, patent-free, NonStop architectures. The desktops would look like the dual-motherboard SGI’s with pluggable CPU’s, etc. Just replace what breaks with the system chugging along using other components.

Somehow I think with the current culture of updates, which is linked to security requirements and ultimately to the fact your computer is always connected to the internet are fundamentally incompatible, thus this kind of computer would need also to be offline, but certainly there are tricks like live kernel updates that could be employed to extend uptime as long as possible.

If average laptop lifetime is about 5 years (for all reasons), then about 0.1% will make it to 50 years and remain operational.

Speculatively, with AI moving along as it is, a “computer” might be very much like a typewriter, primarily a device for creating documents, getting them to peripherals, etc for the human user and their AI ghost/API.

With thermodynamic neural nets, especially if we can get them working at room temperature, we could easily see a situation where it would be more cost and power efficient to have local generative AI simulate standard computer architecture than to actually build a Von Neumann computer from discrete components.

If we can get the thermodynamic wells down to the size of flash cells, that could mean running 1024b models locally on chips the size of an SD card, peaking at around 20 watts at full utilisation.

I could easily see using MCU scale compute to run a stripped down system to provide the wireframe from which the GAI could hang on the pixels and pizaz, helping the user to stay on the rails of a strictly deterministic system while decorating it with GAI. A “50 year computer” might be useful as a stand alone, but basically be an interface device when combined with generative AI running on devices that would probably be much more needful to keep current.

https://www.technologyreview.com/2015/08/06/166822/what-is-t...

The design goal is to build a computer that lasts 50 years. To me this implies a design that is modular and repairable and possibly not based on something you can buy today. I don't want to base my computer on the products that existed 50 years ago or the products I can buy today.

What would I give up in order to get a computer (hardware and software) that lasts 50 years? Size, weight, speed, complexity. Sure.

We now know a lot about change so I need a device that accounts for almost every technology that I use today to have evolved significantly. So I need some long term features.

I want to think in terms of modules, which may be independent physical things. I also want a case to put it all in.

Over the next 50 years I (and my grandchildren) need to be able to repair and replace any part that breaks and continue to evolve the modules that I use, the case and the way the modules interact with each other. My needs will continue to evolve. The rest of the world will continue to evolve around me and I still want to interact with it and its services.

I think some things are constant. I need power; a way to input data; process and store it; usefully share it with others; and a way to output that data.

My modules may therefore include a keyboard, some sort of pointing device and potentially other input devices in the future; a power supply; a bunch of CPU's for various purposes in one or more modules; a set of storage and archive devices; networking; one or more output devices, perhaps a screen or two.

Perhaps the most important thing is an idea, philosophy and a clear idea of what I want the device to do. The article talks about typewriters which are clear on each of these points. I also like the idea that I will need an emotional investment in whatever I end up with.

If I wanted to experiment today I would start with a bunch of Raspberry Pi's and their kindred microcontrollers. Each of my modules would contain one or more of these devices. I would pick a set of connection standards. I don't know where the idea's go from there but it would be fun to find out!

Talk about a killer app.

https://www.defensenews.com/air/2019/10/17/the-us-nuclear-fo...

This is such a horseshit statement. We change those things because of social pressure, not because they wear out. My mother is still using her first generation iphone SE, eight years later. It still facetimes and texts and watches netflix just like it did in 2016. The Nighthawk R7000 router I bought 11 years ago still isn't fully saturated by my network traffic. I have USB chargers in use that came with phones I bought in 2009. My HP printer/scanner is from 2005 and they still make cartridges for it.

https://en.m.wikipedia.org/wiki/Ship_gun_fire-control_system...