A Modular Operating System

A Modular Operating System Information Processing 68 - North-Holland Publishing Company - Amsterdam 1969, pp548-556

The article.

This was a very interesting piece of work. My experience to then was that elegant software was written in the high-level language, like Algol 60 with which I was very familiar. John Martin's interface was a revelation. Elegant KDF9 machine code for an operating system.

As a research project, it was wonderful for me. However, we (NPL) would really have liked to make it into a practical system. It soon became obvious that this was far too much effort and that it was too late in the life of the KDF9 to consider adapting the existing software into that design framework.

KDF9 had a fundamental problem in the address space of 32K words, and relatively slow memory. With hindsight, I am not even sure the system would work with adequate efficiency. Instead, NPL adopted a system which was a modification of the manufactuer's standard system undertaken at Leeds University.

Experimental work into operating systems at that time used minicomputers, rather than a machine like KDF9 which was more like a main-frame.

Reviewing the paper now, it seems like a good write-up. Ian Pyle's question really worried me - he seemed an intellectual giant compared to me giving my first paper at a large, important conference. However, years later he became a friend and was really quite normal!