Is Collapse OS really a practical proposition?

I must start by saying that I have nothing but respect for Virgil Dupras and the folks behind the Collapse OS project. Fundamentally, we all have the same concerns: the modern computing industry is too complex and non-sustainable, and crucial skills needed to work with simpler computing systems are being lost. The attrition of engineers with low level programming skills and low-level electronics skills are things I have commented on before.

What Collapse OS purports to be is an operating system that can run on scavenged computer parts, based on 1980s technology. It primarily focuses on Z80/8080 hardware. There is also a related project Dusk OS which seems to be doing something similar with old PC hardware. Dusk OS seems to have rather different goals, and I won't be discussing it further here.

Building an operating system for a computer made of scavenged parts seems like a fascinating retrocomputing project to me. It's exactly the kind of thing I'm interested in. But Collapse OS isn't positioned as a retrocomputing project -- it purports to be a practical way to create rudimentary computers after our technological society has collapsed. I don't deny that such a collapse is possible -- it's certainly happened before. That we might, in such an event, have bigger things to worry about than playing Zork is something the Collapse OS folks acknowledge. The key point, if I understand correctly, is that having access to workable computers -- even simple ones -- would give us a survival advantage.

Even allowing that computers would be helpful post-apocalypse, I am far from convinced that a practical computer could be made from scavenged parts, and the absence of an operating system is not the main reason why that is so.

Let's suppose you manage to scavenge the bare bones of a Z80-based computer, from an old games console or cash register. These things were built to last and are electrically simple; there's a good chance of finding working parts, and soldering-iron repairs are practicable. Let's suppose also that you have a way to power the device, and a way to solder and make electrical connections, and a safe place to work, and no immediate threats to survival (and none of these suppositions are entirely trivial).

Now what? You have the guts of a computer, with no general-purpose input or output devices, and no storage. The original equipment -- whatever it was -- probably booted from ROM. If it was sophisticated -- something that once ran CP/M, for example -- the ROM would have loaded some code from a floppy disk into memory, and then executed that. Probably most of the low-level I/O control was in the ROM, because that functionality would have been device-specific.

In our post-apocalyptic state, can we burn a ROM? Or even an EEPROM? The equipment may still exist, but it will need to be operated by -- a computer. This is the main, chicken-and-egg problem I see here: we need a computer to build a computer. But let's suppose that we're in a state where we have a way to burn an EEPROM but no other computers apart from what we can build with our own hands. I can't imagine what sequence of events would leave us here but, to be fair, who knows what the apocalypse will bring?

So what about I/O? A general-purpose computer needs some sort of input and output device. We could (presumably) scavenge keyboards and LCD displays from the same places we scavenged our Z80. The ubiquitous 16x2 dot-matrix LCDs are probably going to linger for millennia -- almost every electronic appliance for thirty years had one, and they're indestructible. They are comparatively easy to program, and the data sheet is widely available on -- oh, wait, no Web. But you could make some sort of display device from old seven-segment displays scavenged from clocks -- very early microcomputers had no better displays than this.

But how do we enter the code that will drive our I/O devices, given that we have -- at present -- no I/O devices? I guess that, if we have a way to burn an EEPROM -- assuming the highly speculative state of affairs to which I alluded earlier -- perhaps we could do this by a process of trial and error: burn the EEPROM, test it, burn it again... until we have at least a way to read a keyboard (and this is easy, because it's just a switch matrix) and write an LED display of some sort.

Assuming we can get past that stage -- and I have no idea how -- what will our hacked-up computer use for storage? Not just storage for the operating system itself -- conceivably that could be in ROM -- but storage for whatever applications we run. There are plenty of junk hard drives in the world, but SD cards are probably a better bet -- if they survive the coronal mass ejection or whatever got us to this state. That's true so long SD cards continue to support the old SPI protocol, which is increasingly in doubt. Writing code (Z80 op-codes in hex, I guess) to operate the SPI interface is probably just about tractable -- I've done this for the Raspberry Pi Pico, so I know what's involved. This protocol is clocked, so it will work with slow computing devices. I can't imagine what would be involved in interfacing, say, a SATA SSD to a Z80.

But how will we get the code for Collapse OS (presumably with modifications) into the storage or ROM? At this stage, it looks as if typing it in as hexadecimal is the only practicable approach. We can't rely on any computer-based storage it might be available on. To be honest, I'm not sure we would be able to rely on printed paper, either, but let's assume we have a way. In the early days of the Z80 I used to program directly in hex op-codes using a keypad -- that's not totally out of the question. It wasn't unusual, back in the day, to bootstrap a computer this way: we would enter a minimum program needed to read a rudimentary operating system from, say, punched tape or cards. Of course, we actually had the punched tape or cards with the code on, and we had already used some other computer to get it there -- the old chicken-and-egg again.

One key design feature of Collapse OS is self hosting. For an operating system to be self-hosting means that it can be maintained and developed using itself as the platform. No operating system starts as self-hosting: it's impossible -- I can't maintain Linux on Linux until there actually is a Linux. Linux has been self-hosting for twenty years or more; I presume that other operating systems are the same. Self-hosting is, indeed, a crucial property for an operating system that has to work in highly straightened circumstances. It takes a lot of effort to make a system self-hosting, and it is to the credit of the Collapse OS folks that they have managed this.

Arguably, once you have just one computer up and running Collapse OS, you can use it to port Collapse OS to other hardware. My concern is whether it's possible to get even one computer running it, given the constraints. Perhaps the expectation is that one computer running Collapse OS will survive the apocalypse, and this can be used to build others. But if we are to postulate such a thing, why not postulate that a modern Linux workstation -- or at least a Raspberry Pi -- will be available? With such a thing, we could easily build an operating system (of our choice) for our scavenged hardware.

The elephant in the sitting room is that there are already operating systems designed for 1980's Z80 hardware. The most well-known, of course, is CP/M. There is a colossal amount of software for CP/M and, while its legal status is rather uncertain, that's unlikely to be a problem when we're living in mud huts again.

The Collapse OS team rejects CP/M as a valid operating system for a scavenged computer, and I have mixed feelings about that. On the one hand, CP/M was never really designed to be maintained on CP/M. Even now, whether CP/M can be self-hosting is not clear. As a minimum there would need to be complete source code. While some original source has been released, it's not particular comprehensible, or up-to-date even by CP/M standards.

On the other hand, there are contemporary, open-source equivalents of CP/M. It is almost certainly possible to make these self-hosting.

But would there be any point? The fact that there are many practical applications for CP/M is only a relevant consideration if we actually have access to them. We don't have print-outs of the source, or any tools to build the source if we did. The binaries are distributed around the cloud, and unlikely to be on any kind of storage device we would be able to access from a post-apocalyptic computer.

So, despite what some sceptics have claimed, the existence of CP/M (and similar) is not a knock-down argument against the value of Collapse OS. My main reservation is that, in practice, I don't think we would be able to build a computer that runs Collapse OS -- or anything without access to other computers.

When we look at the real history of computer development, we can see that even the crude-seeming Z80-based appliances of the 1980s are, in fact, fabulously complex. They are an evolutionary step in the process that started with ENIAC, and went through many stages of refinement before we ended up with the Kaypros and RML 380z's, and the Apple IIs, and all the rest of them. These devices were not hacked up in somebody's basement using a multimeter and a soldering iron -- they were the product of a substantial IT industry, that was already well developed. Z80 stuff only looks primitive to us, because we have single-board computers clocked at gigahertz with gigabytes of RAM.

I don't know if it's possible to short-circuit the development of the microcomputer industry, and go straight from scavenged parts to a working computer. The Collapse OS folks say -- and I have no reason to doubt -- that their code runs on the RC2014, TRS-80, Apple II, and others. That's fine as it goes -- but these are complete computers. I might be able to make a TRS-80 from scavenged parts of other TRS-80s, but I doubt I will be able to make one from old games consoles and burglar alarms. It seems to be that there's a huge gap between 'runs on old computers' and 'runs on scavenged computer parts'. And I really don't know what will fill that gap.

It gives me no pleasure to write any of this. I think Collapse OS is a valid project, in that it promotes simplicity and preserves low-level programming skills. It's certainly interesting, and clever. So what would it take to convince me that it could actually meet its post-apocalyptic design goals?

What would convince me is for somebody to make it run on any 1980s Z80 system without the use of any other complete, working computer. It would need to be done as the project itself proposes: from scratch, in an environment in which we assume that no other working computers exist. I don't know whether it's possible to do that even in theory, let alone in practice.

A question remains whether Collapse OS is a better operating system than CP/M besides its post-apocalyptic pretensions. I can certainly think of theoretical reasons why it might be. But CP/M has a lot of momentum, and a significant base of expertise even now. Unfortunately, the population of CP/M experts is ageing; we're competing with the Grim Reaper here -- and that's a competition we can't win. My concern is that the Grim Reaper may end up taking everybody who knows how to solder or to program in assembly language, and then we could be in serious trouble, even without an apocalypse.