It is only recently that I became aware of the existence of this project -- through Mike Scott, whom I regularly bump into at Ceilidh dances here in Edinburgh (well, I'm not as proficient a dancer as he is).

My first involvement with Edinburgh University came when I joined a syntactic analysis project in the English Language department, on which Harry Whitfield and Paul Bratley also worked. I followed them into the Department of Computer Science. The project was under the aegis of Prof. Angus Mc''''''Intosh, and the principal investigator was James Peter Thorne, who succeeded Angus as professor of English Language. Sadly, Jimmy died in post a number of years ago; I still miss him. He was a superb raconteur, and a great source of gossip (only slightly malicious and embroidered) about University politics and intrigue. Not a little of this, it should be said, related to the setting up and evolution of such entities as the Machine Intelligence Unit, the Departments of Computer Science and Artificial Intelligence, and the Edinburgh Regional Computing  Centre.

I eventually left the Department of Computer Science in 1984, to set up Clan Systems, along with Igor Hansen.

It was good to read the message from Prof. Donald Michie. Donald was one of the key people responsible for putting Edinburgh on the computing map, and his annual Machine Intelligence workshops, which attracted leading figures from around the world, created a real buzz of excitement. I wonder if he is considering writing his memoirs. I am sure they would make fascinating, and controversial, reading.

The unforgettable and much-loved Prof. Sidney Michaelson has been mentioned in earlier posts, but another key figure was Dr G E ('Tommy') Thomas, Director of the Edinburgh Regional Computing Centre (ERCC). Like Sidney and Donald, Tommy had vision on the grand scale, and worked tenaciously, and with considerable success, to put in place and nourish the resources to implement that vision.

To return to the Edinburgh Computing History project, I find it intriguing, touching even, that anyone should be interested in re-visiting some of the old stuff created in Edinburgh.
For me, the period was one of great potential and invention, but in many ways the potential was not realised.There was too much emphasis on the individual virtuoso performance, and not enough on teamwork and stability. I accept my full share of the blame for that.

Leaving that aside, many of the problems which we faced in those days were due to insufficient computer resources, particularly storage. This was especially true of the early single-user systems, such as the PDP8, PDP7/9/15, Interdatas, etc. The reverse-loading technique which I used for the PDP9 IMP compiler would not have been necessary if the machine configuration had included either random-access storage or sufficient immediate memory. (If I remember correctly, early versions of the compiler produced the object-code on paper-tape which was then literally read backwards to load the program. All to fix up the forward references of course. Fanfold paper-tape was quite convenient for this).

I would be quite embarassed to have my early implementations of ECCE exposed to view. By any standards of programming practice, they are appalling -- full of jumps and labels to achieve opportunistic code sharing, for example. But it did allow ECCE to be implemented, and efficiently implemented, on a number of modest hardware configurations.
(On the subject of ECCE, I was interested to read Chris Whitfield's remark about the original implemention of ECCE being by Alan Freeman on the PDP8. Until now, I thought I was the begetter of ECCE, although like everything else it built on what went before. My recollection is that Alan Freeman did indeed implement a context-editor on the PDP8, but it did not have the structure and features of ECCE, such as the use of Regular Expressions for programmatic editing.)

As another example of overcoming hardware deficiencies, I devised HAL (High-Level Assembler) largely because I despaired of being able to produce a compiler for machines like the ICL 7502, which had about the worst machine instruction set that I ever encountered. (Not quite the worst. I think a machine called Arcturus was worse. It had 18-bit instructions and 16-bit data words. How you were supposed to load programs was a mystery. The only other thing I remember about it was that it had a SEX instruction -- Sign EXtend, natch, what did you think).

I have to wonder if there is any merit in resurrecting some of this stuff, except along the lines that George Bernard Shaw used about parents: that they should never hold themselves up to their children as an example, only as a warning.

Finally, I still write the occasional IMP program, to meet an ad hoc computing requirement, such as plotting my blood pressure (rising) or the performance of my share portfolio (falling). This runs on a Clan workstation, designed by Igor Hansen, which is a sort of grandson of the APM machine (aka the Fred machine). I also use a laser printer with a Clan-designed controller, which supports a much-enhanced version of Layout, among other protocols.
 
Warm greetings to all former students and colleagues,
Hamish Dewar.