I wonder, what would be your answer is someone came up to you and said "We need a computer that we can maintain and keep in service for 50 years or more. What should we do?".
Take the IBM route. Write to a VM spec. Hardware comes and goes, but VMs last forever. The System 38's virtual architecture is still used on modern AS/400 machines, so programs written in the late 70s will still work today without needing to lift a finger.
Yeah, that makes quite a lot of sense and is a proven technique. On a side note, did they ever build a System 38 VM for something like VMWare, Virtual Box, etc.?
If you know up front that it needs to last for 50 years, and the requirements are unlikely to change, it's easy to justify the expense of getting lots of spare hardware to be able to handle component failure. That's almost certainly going to be cheaper than trying to port the software forward every decade as technology changes.
Besides that, use open protocols and open source for the whole stack. Use open source hardware too, but don't depend on being able to order new hardware. Processes could change (we could move away from silicon to something else), or file formats for schematics can change so that you can't find anyone to produce the product.
Also helpful would be to target a common VM like the JVM instead of relying on hardware-specific abilities.
This is exactly what you do -- lots of spares. If you're big enough, and have a big enough need, you can also design you own hardware ... although even that has limits, since foundries do eventually retire process nodes.
I've read that the answer to this question is why a large part of our military-industrial complex exists - the US government gave Lockheed and others 50+ year long non-cancellable contracts to build equipment for the government back in WW2 because they couldn't justify the business risk (similar to why Fannie Mae and FMAC were created) and it was deemed in the government's "best interest" to keep these companies afloat in some capacity. Aircraft carriers we don't need, guns that are long obsolete, tanks that the Pentagon doesn't want anymore - the list goes on with the excess manufacturing spending the government seems to be roped into buying because of... "reasons."
I've never read through those contracts myself but I would hope that they're public record if they're so old and important.
That sounds a bit hard to believe in the specifics, but in general I don't think it's particularly controversial that the US government has something of a vested interest in keeping the defense industry from contracting too much during peacetime, and structures contracts and purchasing in such a way that companies are kept around who might otherwise go out of business or get acquired. E.g. I think it's pretty much accepted that the Pentagon has gone out of its way to keep the market for military aircraft from collapsing too much. That's an industry that would be difficult to rebuild once the expertise and tooling is gone. (Personally I was surprised when they let McD-D merge in the late 90s.)
Interestingly, there's case studies in the private sector for doing exactly this sort of thing in cyclical industries. Toyota is famous (within the rarefied air of supply-chain textbooks, anyway) for supposedly buying up its suppliers parts and stockpiling them, or even just paying the suppliers to be idle, so that they don't go out of business (or start retooling for other clients) during lean times and are ready to go when demand returns. If you view the military-industrial complex as basically a "supplier" to the US Government, and view the demand for military equipment somewhat cynically as a cyclical demand curve, the government's behavior makes sense. You don't let your suppliers go out of business if you think you might need them again...
That's really the best way to do it; you have to own (or control) as much of the supply chain as you can. Not really practical if you're a private company who has to compete on thin margins, but not really that hard to imagine for a government.
As far as impracticality arguments: keep in mind that each government that maintains a nuclear arsenal (rogue/client states excepted) already has a completely captive technological supply chain, generally at least at a 1950s level, in order to manufacture the weapons in the first place. So if you wanted to build secure nuclear C2 systems, it would make sense to basically build them with the same degree of security, and using the same presumptively-secure supply chains, that are used to create the weapons themselves.
