Long ago, before I graduated from Helsinki University of Technology, I took a job as a field engineer for one of those early computer installations where a computer with the processing power of a modern wristwatch filled up an entire floor of an office building, cost a million 1960s dollars, and had to be run in three shifts around the clock to recover its costs. We field engineers also worked in shifts. It gave me a lot of free time to work on my thesis; on the other hand, when the computer broke down, rapid action was of the essence: every minute of downtime cost the client a fortune.

Finding the fault was not a problem if it was a permanent one. The training we had been given and the tools we had been issued were up to the task. But most of the faults weren't permanent, they were intermittent. And this is where the gremlins come into the picture.

An intermittent fault is normally the result of a component beginning to break or a faulty connection like a cold solder joint or a loose connector. For some time, before becoming permanent, such a fault will come and go, perhaps depending on temperature, electrical load, or vibrations. There in the computer room, however, the coming and going had a pattern to it.

Faults would crop up and crash the computer, and the operators would alert the field engineer. During night shifts, we slept on an inflatable mattress in our little office room and got used to being awake and in action without delay. But when one of us would report to the computer console to observe the fault, all would be well, and would remain that way until the engineer left again. This was a discernible pattern, agreed to by both operators and engineers. Come on: you know that your TV and your computer work the same way: when your grandkid turns up to fix it, it stays on its best behavior.

At the time, we laughed at it, changed some part that was likely to be causing the problem, and went back to the office with the suspect part, there to look for the fault. Usually this approach worked and life went on.

But just think what would have happened if I, at the tender age of 24, with my whole career ahead of me, had had the sense to marshal all the operators and both of my field engineering colleagues to keep a record of these nuisance calls! I'd have had undeniable statistical proof from over a year of working at that computer center, to the effect that the mere presence of somebody with the skills to fix the problem keeps an intermittent fault from repeating itself. Definite, statistically valid proof. I could have gotten a Ph.D. right there and then, even before completing my M.Sc. thesis. I could have become famous... Was I an idiot, or was I not?? Well, we all seem to agree.

How can this behavior of an inanimate piece of equipment be possible? Clearly, it can't be. A machine has no consciousness. But gremlins do.

Gremlins, on the other hand, would be entirely inconsequential, and unknown to humankind, unless they had the means to make their presence and their actions observable. It's easy to conclude that these means must consist of a measure of psychokinetic energy, i.e., the ability to affect the workings of material objects through mental effort. The energy in question need not be large, considering that we people provide the ideal tool: equipment with unreliable components and connections. On that, often microscopical, level, a gremlin can play around with us, because such a fault, while ever so minuscule in extent, has the potential of disturbing or halting the functioning of a large machine.

Now, look at the matter from the gremlin's point of view. Irritating the operators is a lot of fun, and they can't do anything about the fault. Enter the engineer, and here's someone who knows how to deprive our gremlin of his plaything. So, obviously, you do all you can to keep the fault from reoccurring, because that way, maybe, the engineer will go away and you'll have just a little more time to play around. But no, that nasty engineer took out precisely the one printed circuit board that, at this particular time, had a bad transistor on it. Damn.

The time, back in the sixties, would have been ideal for proving the existence of gremlins. We were on duty, not on call. We were only fifteen paces away from the computer console, and every fault was addressed immediately. After those initial years of insanely expensive but unreliable computers, field engineering would never be the same: the equipment became more reliable and had redundancy built in; preventive maintenance kept it running smoothly, and repair calls were handled by whoever was on duty in a geographical area, not by permanent teams of on-site engineers. As the hardware became more reliable, the software became more complex and was soon the cause of most interruptions. Software bugs are due to faulty logic and can be reproduced, but don't lend themselves to playing around by gremlins with psychokinetic powers. I doubt that anyone will ever have such an opportunity as I had then. Woe is me: I didn't think of the chance I had.

I made a lot of typos as I wrote this. Maybe it's gremlins at work?