What if computers went back to the '70s too?

(An interesting article from The Register’s Guy Kewney)

A Wizard whispers in your ear: “The password of Sheffield Library Packet Switching Service is ABC1234XYZ.”

That would be a conversation thirty years ago, on the Multi-User Dungeon, or MUD.

The Dungeon was actually a minicomputer at the University of Essex. I won’t tell you the name of the wizard, because he’s a big noise in the computer business these days, and not a hacker at all, dearie me no. But this isn’t the story of MUD and hackers (we can tell that tale another day). Rather, it’s the tale of how University of Essex came to run a minicomputer.

We have to go back to the early 1970s, and the three-day week in Britain, when there was a crash almost as big as the credit crunch today, and not enough electricity to keep every part of the country warm.

In the Seventies crash, what you wanted was to live near either a computer, or a hospital. If, for example, you lived near Imperial College Computer Centre in South Kensington, there were no power cuts for you. Same if you were near a hospital. Otherwise, power cuts were the order of the three-day week.

But it had to be a real computer, and increasingly, they weren’t. Instead, something called a “minicomputer” was quite likely to be installed in the company data centre. When the power went off, so did it, because it wasn’t important. At least, not in the early 70s.

By the 80s, those minis had started to become the backbone of the equivalent of the internet for those days, and hacking into them was the hobby of an awful lot of today’s captains of industry, who were spotty oiks in those days.

The mystery of why they weren’t computers is one of the weirder stories of those days.

One of the biggest, most successful minicomputer companies was Digital Equipment Corporation, or DEC. It didn’t make computers. The PDP-1 was a “Programmable Data Processor” and DEC was not the Digital Computer Company, because the founders wanted money.

Computers were a dodgy business, for everybody except the one real computer company. All the other seven wannabes didn’t count. IBM was the giant – it was known as Snow White, and companies like Univac, Burroughs, NCR, Control Data Corporation, General Electric, RCA and Honeywell made up the dwarfs.

If that sounds unlikely, you might like to remind yourself that even in the early 1980s, almost nobody knew what a computer actually was. A few who did, still doubted they were either necessary, or good for the world. And so total was IBM’s control of the market that asking for capital to start a computer company would simply make your backers run away. Fast.

So DEC didn’t call its products computers. They were data processing devices and those were, obviously, different. And what they were not was personal. Affordable, but only for reasonably expensive values of “affordable” – only companies could buy even the cheapest Philips Electrological Visible Record Computer or VRC. (It had all the power of a pocket programmable calculator. Something with that little program capacity today would not be able to function as a mobile phone.)

Talk to 70s hackers today and they’ll go misty-eyed about a family of machines from Prime (all forgotten today). They were linked together in a global network which hosted the knowledge index (or dialog) and they were wide, wide open. And why not? The only way to get access to any computer required two things: a modem, and know-how.

Nobody knew what a modem was, and it was almost impossible to buy one. I didn’t get my first until 1979. I was actually accused of breaking the law when I connected it to the phone line. And at 300 bits per second, and no error correction, reading the online manual of Prime computers was a challenge.

Finding out the password, of course, wasn’t any sort of challenge. They never ever set the password to anything. I don’t mean “anything other than default” – I mean they didn’t actually have a password.

What changed all this had its roots in the early austerity years of the 70s – the high cost of computing.

The biggest computer companies in the UK were not the manufacturers, but the bureaux. They bought the biggest iron they could afford, and installed giant mainframes (with roughly the power of a modern iPhone) in big, cyber-scifi offices. You took your program around to them and they put the cards into the hopper, and some hours later, gave you a stack of printout… the company payroll!

But it wasn’t an efficient use of DP capacity.

Just how inefficient that was can be judged by the amount of processor power that went into charging the clients. One giant bureau chief told me in the mid 70s that probably 95 per cent of the processor was used on billing algorithms, so the incentive for a medium sized company to buy its own minicomputer and use it entirely for doing real work was great, and Digital Equipment flourished.

The legendary scientific box was the PDP-8 – small enough to put on a trolley and wheel from lab to lab. You fed your paper tape into the sprocket, and then initiated the run, and carefully stored the results on another paper tape. Nobody dreamed of using those things for commercial apps. But you didn’t need to have an office next to a hospital to afford the electricity, and so when the PDP-8’s big brother the PDP-11 arrived, people did start wondering if these “programmable data processors” might be useable as, well, computers?

What’s odd, is that the question hadn’t been asked before. After all, not only Honeywell and imitators like Prime, but IBM itself actually built and sold minicomputers from the mid-60s (when IBM’s 1130 appeared) and specifically provided standard computing abilities. And, probably, the key is in the price.

When DEC launched the PDP-8, it cost as little as $16,000 (and remember, the dollar was worth nearly three pounds sterling in those days) while the 1130 cost upwards of $33,000. In the early 70s big iron was getting bigger and more expensive and little iron was much more affordable by cash-strapped companies.

What probably made the difference was the PDP-11 and, very soon after it was launched, a development with an impact very like the later arrival of Visual Basic: something called RSTS. Resource Sharing Time Sharing was an operating system which allowed relatively untrained programmers to load Basic programs, and run them in their own time-sharing portion of the PDP-11.

Professional programmers liked the RSX-11 operating system, but you could stick RSTS on a DEC box and just let amateurs have a play, over a dumb terminal. Inelegant, perhaps, but no need to wait for programming staff to get round to your request. The same advantage gave us the PC boom in the early 80s, of course. But there you could actually buy your own system, disks and printer and all, without waiting for the go-ahead of the DP manager. DEC-11s were still far outside the budget of any individual manager.

The mid-70s boom in PDP-11s caught the industry by surprise. The recession probably eased faster than people appreciated – the improvement was concealed behind catastrophes like the devaluation of the pound sterling and several short-term boom and bust cycles. But also the world really hadn’t got a clue what computers were for before 1975. By 1980, everybody had caught on.

The results were catastrophic for one company: Texas Instruments, though at the time it didn’t seem like a catastrophe. What happened was that DEC had hopelessly underestimated the demand for the PDP-11 range, and waiting times went way out into the future. At one stage, companies like Systime were forced to start building their own DEC-compatible memory and storage devices, simply so as to be able to ship systems, because the waiting list (officially) was three years.

Nobody could put a project on hold that long, and so rival companies thrived.

Data General burst out of DEC like a moth out of a chrysalis in 1968, launching the Nova in 1969, and without that initiative to spur DEC on it might be that the PDP-11 wouldn’t have seen daylight. Edson deCastro left DEC when DEC founder Olsen turned down the fanciful concept of a 16-bit architecture. The Nova was so successful that DEC rapidly reversed its decision and definitely beat Data General at the 16-bit game. And DG nonetheless thrived, launching its SuperNova.

Texas’s mistake was to launch a minicomputer called the 990 which was just about good enough to sell as a general purpose computer. Its even worse mistake was to launch a 16-bit microprocessor chip based on the 99 architecture: the 9900.

The 9900 was the first true 16-bit microprocessor and the company planned a secret range of personal computers based on it – a range which would without doubt have dominated the market. The reason it didn’t was simple: it was never launched. And the reason it was not launched was stupid: it was seen as a rival to the success of the 990 mini.

The main problem with the 9900 chip was price: it was loads cheaper than the 990 systems. Excellent! And the idea was greeted by those of us who saw the secret designs for genuine home computers and personal office computers with enthusiasm, because the market genuinely wanted something like this. But the three-year shortage of DEC PDP-11 models meant that TI was shipping every 990 machine it could make.

The board decided to scrap the two top-end personal computers and limit itself to a cut-down 99/4 toy, deliberately hobbled to ensure that nobody could ever try to run an office system with it.

The growth of minis wasn’t limited to “fake servers”, however. Comms was exploding just as hard, and a Honeywell clone-maker, Prime, was poised to explode with it. In the UK, a similar boom with Prestel was about to inflate GEC and its 4000 family of minis.

You might like to speculate, as we say farewell to the genuine minis of the 70s, about just what British computing might have become if the minis that side of the Pond could have been used for more world-wide comms applications, and if BT hadn’t insisted on developing its own System X switching computers (unsellable to any other country). It would probably be dreaming, because GEC, Plessey and Ferranti never showed any interest in general purpose computing. But it’s an interesting dream, all the same.

Source : The Register

Leave a Reply