We move on to about 1975. Several years after the demise of our Elliott 803 machines, and after a period during which our little nucleus of 3 computer maniacs joined, formed and wound up several different companies, we were together again on an accounting project which was to be our main earner over a further five years. We started this project using rented time on the local council's Honeywell. We paid for the time out of the money we earned, and sometimes there was no money left for our salaries. It was a lean time, but our leader, who had considerable experience in ledgers and accounting, kept us going by his expertise and good forward-planning.
Carrying boxes of cards to and from the County Hall, reels of magnetic tape, printer output: and above all, the long waits for our results - for ours had low priority compared with Borough rates and the like - all this was exhausting and often frustrating. Operators had a charming habit of disregarding the run-sheets we filled out with instructions. More than once they mounted the wrong tape on the wrong drive, then over-rode the error messages which appeared. If we specified that a write-protect ring was to be fitted to a tape spool, this was sometimes omitted: and an operator would see "WRONG TAPE ON DRIVE 3", but press an over-ride button to continue! No matter what precautions we built in, there would be some kind of cock-up at least once a month, and output would be rubbish. Backup tapes were of course the only protection, and a re-run would be necessary. We had to keep photo-copies of our run-sheets, since even those got lost or used as scrap paper!
I don't knock the operators, who were under great pressure. The fact was that our work had a low priority, and didn't really fit into their schedule too well.
So we looked around for a second-hand computer we could afford. We had done some programming for a computer maintenance firm: and they heard of this Molecular 18 in a warehouse in Paris. It was nearly new, and hardly used. Six months later the necessary paperwork and re-import licence were completed, and the beastie arrived at our premises.
Money was raised by a bank loan, and for 7,000 pounds (plus transport etc.) we got the processor; one hard disk drive which held an exchangeable single-plate disk cartridge and a fixed disk on the same spindle; two or three more disk packs; a VDU control console; a card reader; and a small Centronics line printer.
For a further sum which I forget, we also got a larger 'barrel' line printer and the special buffered interface it needed. This was a bargain and allowed us to do long printing runs over-night with very little trouble. But we couldn't afford a stacker or decollator (to separate the carbon copies): so we had to take turns on the night shift baby-sitting with the printer. This involved tidying the paper output at intervals; removing this every so often and decollating by hand. You could tell from the sound it made whether anything was amiss, while you ate your sandwiches, did your cross-word or a bit of urgent programming. I used some of these night shifts to work through a disassembly of the operating system.
The processor, disk-drive and power units were in 3 matching cabinets about 4ft high, 2ft wide and 3ft deep. Much of the power cabinet was empty, while the bottom half of the disk unit had a large drawer which could hold 2 or 3 disk packs. We called this the oven, and in fact it kept disks warm and ready for use. After a while we found this made no difference, and stopped using it.
The memory size was 32K, on several plug-in boards some of which were 2K and others 4K. The edge connectors gave a bit of trouble and had to be cleaned now and then.
The 3 cabinets were WATER COOLED! Inside each one was a sort of car radiator with a fan. They were plumbed in, using high-pressure rubber hoses to connect them to copper piping along the wall behind. This was an outside wall, and in the car-park outside was the circulating pump, electrically driven, which also heated the water in winter and cooled it in summer.
This pump would fail about twice a year, and a red light would come on at the switch inside the room. We then had about half an hour in summer to do something about it before the computer overheated. (In winter we could just take off the cabinet sides and continue running.) So we would put "Plan B" into action. We turned a couple of stop-cocks and ran a garden hose from the gents toilet down the passage and into the computer. The outlet went through the wall and down a drain. On one occasion a central heating pipe came apart while we were all out to lunch. We returned to find a flood, and the room full of steam. My boss remarked, "I always thought it was steam-driven!". Happily our new stock of paper had been stacked on shelves.
The Molecular was built in England by Business Computers Ltd of Brighton. It really was an outstanding design which deserved better recognition. I don't know when the first ones were made: I suppose about 1970 or so. But the firm had financial problems and eventually went under. Not many machines were sold.
Its best feature was its incredible multi-tasking operating system. Again, I don't know how many jobs it could do at the same time: more than 6 for sure. I did a disassembly of the operating system and learned how - at each hardware interrupt - all details of the task in hand were stacked; and the job list was scanned, using a priority system to determine which task to resume. The limit was probably the size of the stack area. Factors determining priority included the priority allotted by the operator, how long since a task last had a bite of the cherry, what type of input/output device was involved etc. The control console always had top priority and received instant attention. This was followed by the disk controller which might be completing a "seek" and wanting its buffer emptying or filling. We often had 3 tasks running together - a long printing run from disk, through a formatting program to the printer interface: a programmer doing editing or assemblies: and up to 3 clients putting in data from a network, via modems. The clients shared one program and one data input file on disk. Data was validated as it came in, and tagged with the user ID. Each evening it was processed.
We gave clients a high priority and the programmer a low one: but this seemed to make little difference. We did find that the clients must have higher priority than the printing program, or they noticed degradation in service.
Programming was all in assembler, which was a very nice "language" to use. Particularly useful were the two accumulators A and B which meant less juggling with your variables. Also there was an excellent and versatile file handling library. Several file types were fully supported, including direct access, sequential, indexed-sequential, and a wonderful idea - the "Pool" file on which several users could do their own thing in separate partitions. You could say (program) "give me a free record", or "I've finished with that one, return it to the pool". When you used and wrote a record, you formed your own linked lists and you had to maintain your own pointers and index. But indexed-sequential was a dream, with indexing all done for you to your own specification.
The software manuals were very good, with plenty of examples: and there was a good range of utility software, though of course we wrote most of our own subroutine library. There was an assembly linker which made it easy to link program modules together.
File dumps, diagnostics and memory dumps were always printed out in octal which one quickly became used to reading. Octal also gives one a good mental picture of actual binary values. Naturally the company bought a TI programmer's calculator, which I still have - as one of the three share-holders and directors of the company when it wound up! We shared out perks of that kind; and I also won a swivel chair and my present sitting-room carpet.