The Long Computer

How long can a computer last?

There’s a few extant working examples from the dawn of the computer age, say the 1940s. And while not currently very old, there’s a working model of (part of) Babbage’s difference engine (a mechanical computer) that looks like it could last a while (but also looks pretty hard to fix if something breaks).

The Antikythera Mechanism is in rough shape, but a modern copy, made out of tougher materials, could probably last a while – but it isn’t a general purpose computer either.

The desktop or laptop computer (or phone) that you’re using probably won’t last ten years (15 or 20 if you’re lucky). Good luck replacing old motherboards (or even memory chips). Even better luck replacing that old pre-IDE non-standard hard drive controller.

Minicomputers and other machines aimed at big businesses perhaps will hang around a while longer, as parts are kept in stock, and the underlying software is intended to be stable over long periods of time. I’ve occasionally done work on machines of this nature from the 1970s, and they’re often barely still functioning though. Even if the hardware still has some life, there’s few people left who understand how to operate them.

So, what would happen if something bad happened to global supply chains, if our current global civilization somehow falls? Having working computers would be useful in fixing things, right? But how do you maintain a device where the chip comes out of a multi-billion dollar fab, and the components are manufactured in a dozen countries, assembled in vast factories, and powered by highly specific voltages?

Would it be possible to build a general purpose computer that can last for hundreds, possibly thousands of years? That would be easily maintainable, even in low-tech scenarios?

I’m somewhat inspired by the guy that built a stone age telegraph. It’s pretty clear that with the right set of knowledge, simple electronic devices are, well, simple.

I’m also thinking of the Long Now Foundation, and their Ten Thousand Year Clock. Clearly it is possible to design devices that are intended to last a long time, possibly under adverse physical conditions.

In second year computer science (I won’t say how long ago), I did a course in chip design. Starting from basic principles, we designed a simple RISC chip that had roughly 60,000 gates (i.e. transisters). Nothing fancy, and extremely slow by today’s computational standards. However, the simple and modular design would lend itself well to a device that was made using physical electro-mechanical relays (basically fancy light switches).

Could such a device be designed so that each of its components could easily be removed and replaced, and those replacement components could readily be built out of crude materials? Could such a device be designed for tolerance of a wide variety of voltages and physical conditions (heat, cold, moisture, dust)? How could the operating system and software be designed to function over centuries (and linguistic shifts)?

With appropriate maintenance, how long could such a machine survive for?