As a teenager, I spent hours reading datasheets for CPU components like the bit slice processors AM2901. I also designed hypothetical CPUs in block diagrams. A 16 bit CPU was my objective, and it should have had a clock of 20 MHz. The instruction set was to be small and beautiful. My first computer was based on a National Semiconductor SC/MP processor, which had a very nice and simple instruction set, and I’ve always liked things which are the simple, beautiful and logical.
I never did build my own CPU, but at university, I spent hours playing with the old RC4000 mainframe we had in the electronics club. It was (is!) a 24 bit machine from the late 60′s, a very nice and advanced design. Its concurrent programming system was among the first in the world (appearing at about the same time as Unix was developed in the US), and my father, who used to work at Regnecentralen, gave me one of the original books about the system, which I read with great interest. It was a message passing based system, and I think the design had a few drawbacks, but the beauty of it was fascinating.
Yesterday I saw my old friend again. She is now residing in the rooms with computer historic enthusiasts. I visited the club with 14 year old son Frederik.
The RC4000 still works, but the enthusiasts are more focused now on its predecessor: A GIER computer, which has recently been brought back to life. The first GIER was delivered on July 31st 1961, so the design is turning 50 years old this year. About 50 of these machines were built and the club owns two: One early model and a late model. It is the late model which they are bringing back to life.
There are very few test programs for GIER, and this makes the repair process a bit more complicated. Poul-Henning Kamp, the author of the Varnish Cache, who is one of the active enthusiasts in the club, mentioned that it was probably because the designers found test program development to be too boring. These guys were innovators! Poul-Henning has unit tests for Varnish with great code coverage, but admits that writing tests is not the fun part of that project.
I used to program for a living and I can agree with this. Unit testing isn’t fun!
The smell of the old computers, their physical size, the noise they make, and their sheer lack of computing power (by todays standards) suggest that these machines are of a different era. Technology has evolved a lot, but people are still people. And the people matter – the technology is just a means to an end.
I still love working hands-on with technology, but my view on IT has grown a lot wider since I was young.