As promised in the previous issue of ECN, this final issue concentrates on looking back over the 20 years that the Rutherford Appleton Laboratory has been involved in the support of computing for the engineering community. We have tried to cover most if not all aspects at some level, although inevitably we have not been able to cover everything in great detail. I am most grateful to Bob Hopgood, Paul Bryant, Dave Lomas, Ken Robinson and Bill Swindells for their contributions to what I hope you will find to be an interesting final chapter in the 20 year story.
I am pleased to be able to finish on a positive note (see David Boyd's article on this page for the details) concerning the 3 Community Clubs. Although the EASE Programme officially ends on 31 March 1996, these 3 Clubs will continue to operate as EPSRC Community Clubs for at least the next two years with some central funding support from EPSRC.
As you will now know, EPSRC has agreed to continue its support for the 3 Community Clubs for a further 2 years with a reducing funding profile. This is partly in response to the strong support for the existence of the Clubs expressed in the recent survey of grant holders. However, the survey also showed a lower than expected awareness of the Clubs among this community.
One of our highest priorities will therefore be to promote a greater knowledge of the Clubs and their programmes. To this end, we will be starting a new 4-page newsletter to carry information specifically about Club activities. The first issue of CC news will appear in April. This will be sent to all current EPSRC grant holders and registered Club members with a reply slip inviting registration for future issues. With the closure of ECN, this will now be our main way of reaching the community so we encourage you to respond to this invitation. We will also be making information about the Clubs available online using the WWW. Details will appear in CC news.
One of the major objectives of the Community Clubs is to disseminate information of benefit to their members. To this end, we will be setting up an online information service which will make available to Club members a wide range of material relevant to their research interests. Our intention is that this should encourage the sharing of information of common interest within the research community. Membership of the Clubs is of course open to industrialists as well as academics. The Clubs therefore also have a potential role in transferring research results to UK industry through their programmes of seminars and workshops. We have been looking at ways in which we can make the operation of the Clubs more efficient and economical. Much of our organisational effort will therefore be common to all 3 Clubs. However we will still ensure that each Club remains sharply focused on the needs of its members. To safeguard this, the present Steering Groups will continue to advise us on the programme for each Club.
While the Clubs originally grew out of the engineering sector within EPSRC, it emerged from the survey that many of their activities are of interest to a wider constituency within the EPSRC research community. We will respond to this as far as resources will allow. Our goal is to make the Community Clubs better serve the needs of the broader EPSRC research community and, through their activities, to improve the quality and efficiency of EPSRC's research programmes.
I am responsible for the overall management of the Community Clubs programme co-ordinated by RAL on behalf of EPSRC. I would like to hear from anyone who has views on how the Clubs could better serve the research community. These are your Community Clubs. Please support them and benefit from them.
(Extract taken directly from the original report was reprinted: Sections 3, Table 6.5 and Table 6.9)
The GEC 4080 computer was designed by Mike Mellier Smith in the early 1970s for real time applications. By the time the Rosenbrock Report was published in 1976 and the search for a multi-user mini to put in engineering sites was on, the 4080 had sprouted a multi-user operating system - OS4000. At that time the buy British policy was still alive and there was still much enthusiasm in high places to fly the flag. For my part, I was still in love with the machine having worked with it for the 1906A front end project.
To everyone's surprise, the 4080 performed well against its competitors. Perhaps the fight was not as fair as it should have been since Dave Toll (another 4080 lover) and myself put in much effort to tune the 4080 and, to coin a phrase, could cut a mean program when it came to the GEC.
Dave Toll and myself had a vision. The Rosenbrock Report said the multi-user minis should be networked. Our vision was to really network the 4080 computers. The first aim was to implement the Coloured Book protocols and to extend the emerging SRC X.25 network and connect the 4080s to it. The second aim was to manage the machines from Rutherford. By manage we aimed to support the operating system software from Rutherford and not allow the sites to fiddle about. We would thus be able to update the software from Rutherford with almost 100% certainty that it would work. We could also log into the machines and poke about to chase faults. We thought that this would give the most economic and best service.
Luck was on our side. As well as Dave Toll, I was exceedingly lucky to have Philip Gladstone, Jonathan Mills, Lesley French, and Shirley Wood who were all hardworking and talented 4080 experts. Jacky Hutchinson was very able as user support and not to be forgotten, Roland Brandwood and Liz Krauesslar on operations.
The worst problem was delivery. It took 6 months and more to get delivery. The first machine went in at Bristol in October 1978. It turned out that we just managed to get the communications software finished in time. From that date all new software, including the operating system, was delivered via the network. Transfers were usually done at week-ends as they took so long over our slow lines and on Monday morning we checked to make sure everything had gone well before getting the local managers to re-IPL the machines. This was not really necessary but a precaution just in case it all went wrong.
If my memory serves me right, there were eventually 30 or so 4080s or developments from it attached to the network. Also, as far as I remember there were no major disasters. Much of the credit must go to the superb team. The software support was of a high quality. The user support was second to none.
In installing 4080s we took great care, starting with the preparation of the computer rooms and some were of more than passing interest. In Glasgow the machine was underneath a chemistry laboratory and noxious chemicals dripped through the ceiling from time to time. The GEC was therefore placed under what looked like a car port. This machine was also run by an all female staff - Ann, Anne and Linda. The Leicester Poly (now grandly known as De Montfort University) was the first machine to suffer from asbestosis and was entombed for some time before being rescued through the window. We must not forget Cambridge who had a very mobile machine which had network contact but could not be found - it had somehow moved by forces mysterious.
Each 4080 had a manager and managers meetings were held every 6 months or so. The Glasgow meeting was one to remember. It was Jacky who calculated that it would be cheaper to hire a plane from Kidlington for the group to get to Glasgow than to take the shuttle or train. And so we did.
Management went ballistic. It did not matter that it was cheap but the issue was what would the neighbours think. This episode is still talked of in hushed tones.
As far as I know this was the first occasion when computers had been controlled over a network. Also as far as I am aware the last. Both the PRIMEs and the Starlink VAXs, although networked, expected the local managers to support the operating system software. To my mind we developed a very cost effective means of supporting computers that was never developed. As an aside, life is getting worse with just about every PC user being an amateur systems expert if that is not a contradiction in terms.
The 4080 computers and later the PRIMEs turned the infant X.25 SRC network into a quite extensive and very functional network. It proved beyond doubt that the concepts behind the Coloured Book protocols actually worked rather than collecting dust on shelves. It is my belief that without SRCnet we may well not have had the JANET network or at least it would have been much later in time.
The GEC 4000 series only had modest upgrades and eventually was eclipsed by the DEC VAX. In retrospect it was perhaps inevitable that a small British manufacturer could not compete with a large American company for any length of time.
I never really found out just how useful the 4080s were. Certainly people used them. Certainly we had few complaints. Certainly there was often some initial tension between computing departments and multi-user minis. There was a thought that SRC was taking everything over with generous funding whereas the computing departments were hard up. On the other hand the recipient departments were often frustrated in not receiving the resources they needed. Remember this was in the age when computing departments were still the holy of holies - that was well before Bill Gates changed all that. It was an exciting project where for once resources and circumstances resulted in innovation and services.
The first modem to pass the portals of the Rutherford Laboratory (RL) was a two foot square monster courtesy of the General Post Office. It delivered an amazing 2,400 bits per second. The leased line it was connected took six months on bended knee to achieve. That was circa 1970.
During the early 1970s Rutherford, Atlas, and Daresbury Laboratories all started putting card reader/line printer work stations into Universities. This replaced the sending of jobs to the central machines by post or courier services.
The real start of networking can be accurately dated to 22 March 1974. This was when the first of many meetings were held as the result of a proposal initiated by Trevor Daniels of Daresbury and John Burren of Rutherford to network the two organisations. Atlas, then a separate laboratory, also joined the consortium. The reasons for the initiative were obvious. Some sites had two or even three workstations to the central sites and that was wasteful.
1974 was memorable for a second reason as it was that year that the Post Office launched the Experimental Packet Switched Network. John Burren's idea was to use the EPSS protocols for what was soon named SRCnet. The project was to use the IBM360/195 as a packet switch at Rutherford. Daresbury decided to use a PDP11 and Atlas decided to front end the ICL 1906A with a GEC4080. In retrospect a little more economy could have been possible with a little more commonality of equipment.
In 1979 the Joint Network Team was set up under Roland Rosner with the aim of co-ordinating networking and perhaps setting up a national network of some sort. Their main activities were pouring resources into networking with the aim of keeping development on the straight and narrow - otherwise known as X.25, Coloured Book protocols and Cambridge rings, a far sighted initiative.
The emergence of X.25 meant that the EPSS protocols had little change of life. The decision taken in 1979 was to abandon EPSS SRCnet and to turn the GEC 4080 into an X.25 switch.
Daresbury also decided to write switch software but for a PDP11. And so in a fairly short time there was a small X.25 network with two switches, and GEC 2050 and PDP11 workstations. The high level protocols used with EPSS were HASP and the EPSS Interactive Terminal Protocol (ITP) which were preserved in the change.
Things would probably have gently drifted. However, the 1975 Rosenbrock Report wanted the interactive machines they proposed to be networked. This coupled with the emergence of the SRC X.25 network (probably the first such network in the world), the just emerging network enthusiasm plus the team of network junkies in charge of putting GEC4080 minis into Universities, all came together just at the right time in 1980.
The GEC multi-user mini team had a head start with the X25 code already written for the switch. This was easily adapted for a multi-user mini. Surprisingly, the production of the first two and most important protocols was easier than expected. Philip Gladstone implemented Blue Book File Transfer in about 9 months and I produced ITP in about two months. HASP was also provided by Gill Walters of Particle Physics. Its importance came later when we attempted to draw together the multi-user mini Rosenbrock clients with the High Energy Physicists.
Being first has its problems. We needed a file transfer protocol. The first JNT workshop took place in Glasgow in 1977. The gurus were muttering about yet another revision of the emerging Blue Book FTP. My main claim to fame was that I warned the conference that I was starting implementation within the month and I was not going to reimplement if they decided on revision after revision. The result, I like to think, was that Blue Book was published.
It was a surprise to us that the communications systems we constructed were very robust. How it was used is in the companion article The last of the British minis. We invented network status protocol. This allowed us to interrogate what was happening on a remote machine. Indeed. it was an early sort of SNMP.
By 1979 the Natural Environment Research Council (NERC) had joined the consortium and York University was a fellow traveller as they were involved with the networking of DEC machines.
In 1981 we recognised that electronic mail was needed. Fortunately, the community decided to adopt the ARPA RFC 822 and so we were spared the pain of a UK parochial offering which would have caused unending problems when connections were made to foreign networks. I have to admit that I attended the infamous meeting in London when the UK mail domain order was decided. At the time ARPA had not decided on an order. The arguments were all about how one wrote addresses on envelopes and the order of telephone numbers and area codes. The wrong one won - and don't we know! I apologise for having supported envelope order. The June 1982 SRCnet meeting showed the list participants for the first time with their email addresses.
The final Coloured Book was red (job transfer and manipulation protocol). At the Cambridge Networkshop in 1981 the call went up for a simple job transfer protocol. John Larmouth obliged with Red Book. The word simple may be an exaggeration as Red Book was pretty complex and, to quote Dave Toll, did just about everything including a chorus of land of hope and glory. Lesley French implemented Red Book on the 4080 and this was probably one of the most demanding protocol jobs. It is debatable whether it was worth the effort as in the event it never became popular. Salford University produced a portable version of the protocol written in FORTRAN which we were pressured to use but resisted. It turned out in trying to implement this on the Rutherford IBM that porting it took more effort than reimplementing it. Even then the portable version lacked many facilities. But let's be fair, it was an impressive and well designed protocol.
The PRIME computers started to be delivered in 1979 and came with X.25, Blue Book File transfer and triple X code. But the sting in the tail was that they used HDLC and not bi-synch. The upshot was that they formed a small network of their own as they were capable of switching. Eventually bi-synch was implemented. But, interworking with the GECs was not easy. It turned out that the PRIME products were rather minimal and really designed for PRIMEs to talk to PRIMEs. So some effort was needed to get the machines to interwork.
One of our problems was our use of bi-synch protocols. GEC was quite late in bringing out X.25 hardware. It took several years before we finally phased out bi-synch after HDLC started to arrive in 1982.
Meantime Daresbury had not been idle. Their switches based on PDP11s had matured and formed an integral part of the SRC network. NERC was impressed and set up a network based on these switches which was connected to the SRC one. This was important to NERC as they had a large number of research stations spread round the UK. Thus by 1981 we had a network of about 30 nodes. We had invented an X.25 address scheme. Rutherford machines started at 1 and Daresbury had some other range. This was not a good idea since an X.25 address is 15 digits including 2 sub-address digits. This eventually led to some interesting questions as to whether the address 1 was 1, 100, 000000000000100, or 000000000000001. For those with X.25 expertise we would have been well advised to have given ourselves a DNIC (the first 4 digits of the address with the leading one certainly not zero).
The DEC VAX was now on the scene. Various groups were implementing Coloured Book protocols. Indeed, charts were starting to appear of the progress of each protocol on each machine. There was much rivalry between the various teams. I like to think, as the leader of the GEC team, that we were at the forefront.
1982 saw the emergence of GEC's own products. It has never ceased to amaze me why GEC did not adopt the RL implementations. This would have saved them a lot of time and money. Although we eventually adopted their products, this was to allow us to reduce support costs and not because the GEC offerings were any better.
Meanwhile the Post Office had set up an X.25 network and this was increasingly being connected to networks in other countries. We decided to build a gateway between SRCnet and the Post Office network, known as PSS, that first worked in 1981. Andrew Dunn built the gateway which was principally for interactive traffic. This allowed us to access many machines overseas - in particularly CERN who were getting interested in X.25. The gateway lasted until 1995.
The Post Office then, as BT now, did not like loss leaders. From the first their X.25 service was expensive and under powered. When we were using 9.6K lines they used 2.4K and so on. At a meeting in 1978 BT stated that fast circuit switching was the future and we should wait a bit before setting up a network as next year X.25 would be obsolete. Fast circuit switching or ISDN took a little longer than expected - another 18 years.
From about 1980 the need for a national network started to emerge. JNT had supported several regional networks using various technologies. The questions were should the network be provided by the Post Office or should a private network be set up. If a private network, should it be set up from scratch or developed from an existing network. For starters, the Post Office ruled themselves out by being totally unwilling to do a deal to make their network anything like economic. SERC (as SRC had been renamed) was quite keen to rid itself of its network. It was expensive and was being used by group only loosely connected with RL. Clearly adopting a well developed, extensive network was the way to go. The critical proposal came out in August 1982. And that was the end of SRCnet, even though the take-over did not actually take place until April 1984.
Hindsight is easy. It is interesting to speculate how networking would have developed if SRC had decided to use ARPA technology. Had SRC adopted ARPA, to my mind, the attractions to X.25 and Coloured Books would have been far less. JNT in 1985 would have had to adopt IP or set up a new X.25 network and try and persuade SRC to change - not easy. Therefore, it could be said that the decisions in 1974 delayed the adoption of IP in 1990 by many years. Mea culpa.
This is the story of how we tried to make the UK a major force in computing world-wide. It has all the characteristics of a good novel with a start, middle and dramatic end, not that we knew that when we started. When the Rosenbrock Report was published in 1975. the only multi-user interactive facilities available were on large mainframe systems. In consequence, the main Rosenbrock recommendation of installing multi-user systems was quite adventurous. Even more forward looking though was a statement that past 1978 single user systems on people's desks might be a viable option.
Between 1977 and 1979 RAL tried a few embryonic single user systems (not yet called workstations) but without much success. They were all underpowered and over-priced. Tektronix 4051s and IBM 5100s were two I remember.
The Computer Science Committee of the Engineering Board had set up the Distributed Computing Systems Programme (DCS) in 1976. UK researchers had visited Xerox PARC and seen the in-house Alto systems and there was a demand in the UK for similar systems. George Coulouris at QMC was already thinking about the paperless office and the integrated personal information system. They made some attempts to use LSI-11s an PDP11/34s for this purpose but with the same conclusions as the engineers. The devices are just not on the market yet. Gill Ringland and I were the initial coordinators of the DCS Programme with Rob Witty and David Duce taking over later on.
May 1979 was the time when rumours started spreading that a genuine commercial workstation was under development at a small Pittsburgh company called Three Rivers. Mike Rogers, Head of the Computer Science Committee, had visited Carnegie Mellon University and been told about it. George Coulouris received a draft specification and passed it on to Rob Witty. David Duce and I were at the Seillac II Workshop in France where Alan Kay of Xerox PARC, Bob Sproull of CMU and Ron Baecker from Toronto were all enthusiastic. Bob and Ron had letters of intent in to purchase systems.
Three Rivers was a company formed in 1974 by 5 people mainly from the C.mmp project at CMU. They had produced a really fast vector display, a good colour system and had installed colour bank teller systems in Mellon Bank branches in Pittsburgh. Brian Rosen, one of the founders, left in 1976 to do product development of the Alto at Xerox PARC. Xerox turned out to be over cautious and he rejoined Three Rivers in 1978 and this is when the company committed itself to marketing the first commercial workstation. By now the company was about 10 strong.
The product initially had the name System Q, then Oakland, Pascal Alto and finally PERQ from perquisite which has a slightly different meaning in the USA having more the flavour of added-value than something given away to an employee! It does not normally get abbreviated to perq or perk as it does in the UK. It was to be a 1 MIP system with Pascal P-Codes as the order code, 256 Kbytes of memory (which is all we had initially on the 6-user GEC systems), and a 12 Mbyte disc. The most impressive part was a 768 by 1024 bit-mapped display with exceptional performance. Stan Kriz and Brian Rosen knew how to make displays and continued making a living in this area way after the collapse of Three Rivers many years later. A high resolution display for radiographers is currently on the market in the USA that they developed. Initial PERQ systems were to have a touch-sensitive pad similar to the one on the current Apple Powerboats but quite a bit larger.
We were being urged by the community to buy a system for evaluation and after discussions with the Chairmen of the ICF and DCS programmes, I phoned up Three Rivers and said we wanted to buy a system and have an option on a further 10. There were 25 systems being built in the pilot production phase. While several people had given Three Rivers letters of intent, the SRC order on 27 June 1979 was the first they had received. The cost was about £10K. (That was the price we were paying for a Tektronix 4014 storage tube) We had the problem of getting the order to them quickly and the British Embassy in Washington placed the order for us and paid for it when it was finally delivered. (To my knowledge, we have still not repaid the British Embassy for it!) The first prototype was to be demonstrated in August 1979. CMU who were sharing the software development would get some of the early systems and ours was scheduled to be delivered on 30 November 1979.
Tommy Thomas from the Edinburgh Regional Computer Centre visited RAL in July and mentioned a couple of confidential R&D projects being done at ICL's West Gorton plant which had some similarity with the DCS programme. The one that was of immediate interest was a project to put an A4 bit-map display with a Motorola microprocessor to produce a prototype workstation. We arranged a meeting with Charlie Portman which eventually took place in September. Prior to that, I attended SIGGRAPH in Chicago in August and was able to see the prototype of the PERQ system that we had ordered. The Three Rivers stand was engulfed for the full three days of the Conference. Out of the several hundred displays on view, it was by far the most impressive. The A4 display was far superior to Alto's display and everybody was pushing to get in an order.
The visit to West Gorton was rather disappointing as their workstation was in a very early stage of development. However, it did have a grey-scale display rather than monochrome which really improved the quality of the text. We brought out the glossy of the PERQ machine we had ordered and this had a devastating effect. The Three Rivers product was probably two years ahead of the ICL development even with strong backing in the company. SRC that time still had a patriotic buy-British policy and we urged ICL to buy Three Rivers or at the very least come to some arrangement with them.
Getting ICL interested required the senior management to be convinced. In consequence, I wrote to Ninian Eadie in October urging ICL to buy into Three Rivers. It would allow ICL to establish world-wide penetration while developing future systems to exploit the exploding market we saw emerging in the 1980s. This is an important opportunity that should not be missed in establishing ICL as the leading workstation manufacturer.
Ninian Eadie responded positively. They also ordered a PERQ system for evaluation. ICL visited the Three River manufacturing facility in October and a joint meeting took place between ICL and RAL in November where it was agreed that the two PERQs would be evaluated by the hardware team at West Gorton and the software team at RAL. The strategy outlined at that first meeting was not vastly changed throughout the project. There would be a need for a cheaper Mark 2 PERQ no later than 1982 and a true 32-bit system that sat on the desk by 1984 at the latest. Between May and November 1979 a strategy had been put together that looked good. On 6 December, 1979 the Engineering Board received its first grant application for a PERQ from Newcastle University. All seemed to be going well. The only problems were a rather cautious attitude by ICL due to their own financial situation and the non-appearance of the first PERQ that was supposed to arrive in November but had slipped to January due to some hardware faults. Grant applications had started to arrive with SRC in ever increasing numbers. Three Rivers financial state was also not brilliant in that it had to get a $0.5M loan from the First Bank of Boston (arranged by Ed Fredkin, an MIT Professor) to pay the operating costs during the second half of 1979. Ed put in $50K of his own money and got a seat on the Three Rivers Board. When the systems started selling in 1980 the cash flow would improve.
Three main activities continued in parallel during 1980. Three Rivers tried to produce the machines. ICL negotiated with Three Rivers, and SRC attempted to set up a Single User System Programme.
The prototype we had seen at SIGGRAPH had wirewrap boards and PCB versions of these three boards had been subcontracted once the design was complete. The processor board was a 6 layer board with 290 chips and Three Rivers was let down by its supplier so that much of 1980 was spent in solving this. Orders had flowed in from many large companies in the USA including DEC, IBM, Bell Labs, Philips, Olivetti and Carnegie Mellon (the software partners of Three Rivers who were committed to buying at least 10 systems). By February, Three Rivers had 50 orders. To cut a long story short, the first PERQ was shipped in November 1980 rather than November 1979. This was a major slippage but not fatal. It was only in May 1980 that Bill Poduska left PRIME to set up Apollo. There was rumours that a small company called Apple might be building a machine called the Lisa. A graduate student from Stanford had set up a small company to produce a processor board which had rasterop hardware. At the 1980 SIGGRAPH in Seattle, two PERQs were on show with the first commercial overlapping window system (so much for the later law suits between Microsoft and Apple as to who was the originator of the desk top metaphor!).
These delays had hurt Three Rivers. Rumours were around that Apollo would announce a product in the Autumn. I phoned Bill Poduska and he said they hoped to have a prototype by April 1981. It was agreed we would purchase the first system to assess against the PERQ. ICL visited Apollo for discussions and the Apollo launch came in November 1980. However, Three Rivers had started shipping systems to CMU. They had 19 people working on the SPICE project that would use the PERQ. They were developing a microkernel operating system called ACCENT which later developed into the now famous Mach kernel.
The negotiations between Three Rivers and ICL continued throughout the year. We were not totally aware of all that went on but ICL were being fairly cautious in their predicted sales. Internally there were rival products so a low profile had to be maintained and throughout it was labelled as an engineering workstation even though many of our initial demonstrations clearly showed its potential in ICL's office systems and commercial markets. The people in Three Rivers were split between those who saw they could sell a few thousand systems quickly if they could ever make them and so were eager to get at ICL's manufacturing capability and those whose baby it was and they saw it as a potential loss of control. ICL had acquired Singer at Utica and Three Rivers also saw that as a threat. Three Rivers had plans to go for $15M of capital funding and open their own large plant with 300 staff. The situation did not get resolved until Ed Fredkin took over as Chairman and Chief Executive in September 1980. By the end of 1980, ICL's own financial viability became rather less secure and Three Rivers were joking about buying ICL rather than vice versa.
Throughout 1980 meetings took place in the ICF and the DCS Programme to attempt to produce a joint initiative. In February, the DCS Programme agreed to order 6 PERQs with funds available to purchase another 24. ICF purchased 3 (making the initial 10 that SRC ordered) and committed to buy 50. The new Software Technology Initiative that had come out of the Robert's Panel report committed to purchase a further 25. Thus we had commitments to purchase over 100 systems from different parts of the Engineering Board.
The first SRC PERQ was shipped from Three Rivers in late November 1980 and was shipped to Washington for the Embassy to onward route (if you remember they had ordered it for us). It got lost. Three Rivers insisted they had shipped it. The Embassy said it had not arrived. Later it turned out that it had arrived but there was problems getting an export licence. There was some surface damage on the cardboard box that the machine had arrived in so the Embassy decide to re-pack it (the boxes were designed to withstand a fair amount of battering and we never had any trouble with subsequent machines despite often arriving in a rather scuffed state). The Embassy constructed a very sturdy wooden crate taking a great deal of time over this, not seeming to be aware that there was some urgency in the delivery (to be fair it was a year late in delivery). It finally arrived at RAL on 21 January 1981.
I ceremoniously opened the crate with a crowbar provided by the ICL engineers who had come down for the occasion. It took a while as the Embassy had bolted the lid down with 4 inch nails some of which had gone right through the wrapping surrounding the PERQ and into the PERQ itself. Securing screws for both the disc and the card cage had fallen out. One of the boards had a corner almost completely destroyed as the PERQ had bounced up and down in the wooden box. It looked a bit of a wreck and the SRC people saw it as a major disaster while the ICL folks from West Gorton saw it as a challenge. Within a surprisingly short amount of time they had the system working and this system was widely used in demonstrations during 1981 without failing. I kept it around until last year hoping that RAL might be interested in starting a museum but lack of interest finally meant it was cleared away.
Everybody was eager to see the beast so we put together a number of demonstrations including a simulated office based on a multiple window paradigm similar to what later appeared on the Mac. We wrote some programs to demonstrate the graphics power of the system. One moved four quarter-screen size photographs of ICL Directors about the screen. We had a job getting this to work until we finally realised that it was updating the positions so fast that we were not seeing the intermediate positions before the program got back round the loop and displayed the initial positions again. We had to put in artificial delays so that people could see the impressive graphics quality! Over 100 people had demonstrations in the period March to August. Most of the SRC Boards, the other Research Councils, several Government Departments, the ICL Directors, many universities and a number of government officials all visited. We had visitors from many European countries and further a field. It was a strain on our resources especially as the key systems people were often required to do the demonstrations.
The presentation to the ICL Directors went well but the state of ICL was getting very bad. The share prices dropped from 196p in September to 40p in March and their debt increased by around £100M. This was the period when customers were turning away from mainframes, ICL's core business. Keith Joseph announced a £100M loan to ICL on 19 March. The first set of ICL Directors saw the PERQ on 18 March and the second set on 20 March! Ed Fredkin was due in London on 22 March to finalise the negotiations with ICL. Kenneth Baker the Minister for Industry, declined to express his full confidence in ICL's management and implied that management changes might be necessary. Even so, a frame agreement was reached with ICL giving ICL exclusive rights to market in the UK, South Africa and Australia. ICL had to commit to purchasing 200 systems in the first 2 years. Manufacture in the UK would start about a year from the date of the agreement. A presentation was scheduled with Chris Wilson, the head of ICL, in April. It never took place. By May 1981, Chris Laidlaw had been installed as Chairman and Robb Wilmot had been brought in as Managing Director with an immediate goal of doing a thorough review of all ICL's existing projects before announcing ICL's new strategy. We invited Robb Wilmot to come and see the PERQ!
ICL's PERQ arrived as did one for Logica. To ease ICL's financial position a little and also Logica's, who were also having some financial problems, we bought their PERQs off them so RAL had three PERQs bought from three different companies at this stage! There were a number of discussions at this stage concerning a joint ICL/RAL/Logica activity aimed at producing the software on the PERQ necessary for the scientific community. The 3 systems were all slightly different and swapping keyboards, monitors or anything else tended to result in disaster. In consequence, we devised a colour coding system so that all one system had pink labels, another blue and the third yellow. The trouble was we kept coming in each morning with the systems all in a mess. It was a while before we realised that Rob Witty was colour blind and that he was inadvertently screwing things up. I had employed Rob earlier to do software development on a colour microfilm plotter! The one question I had not asked him!
SRC, including Geoff Manning, the Director of RAL, had many discussions with the Department of Industry during this phase. It was agreed that the Single User System programme should be a SRC-wide one and not restricted to the Engineering Board. This was approved by Council in July 1981. A Common Base of software which included Unix was needed for the PERQ and the estimate was that about 16 staff years of effort would be needed in the period 1981/82 to achieve this. To ensure that ICL had a viable cash flow during this period the aim was for SRC to commit to buying 200 systems over the next 2 years.
Negotiations between Three Rivers and ICL continued after Robb Wilmot took over. We demonstrated the PERQ to him one morning in May at 8.30am in Putney. He was impressed and saw its potential as an Office System if the price could be brought down below £10K. Apollo were launching their first machine in the UK on 3 August and it was important that the ICL/Three Rivers deal was completed by then.
The plan was for ICL to conclude the deal with Three Rivers in June, after the Council approval in July to announce the major order from SERC and then in August to announce the SERC/ICL Collaboration to develop the necessary software.
The RAL Apollo system arrived on 24 June. This had a green 30Hz display which was not very impressive but the company had plans for a 60Hz black and white system in 1982.
Needless to say nothing went smooth. The agreement with Three Rivers got bogged down in that ICL wanted a commitment from Three Rivers that they would be the first company to produce a system under $10K. There was a complete reorganisation in ICL on 16 July, the day after the Council meeting. Bill Poduska arrived in the UK for the Apollo launch and gave a presentation at RAL on 31 July. I was due to go to the USA on 10 August for visits to IBM, Three Rivers and Apollo. Apollo were interested in a deal with ICL. By the time I got on the plane at Heathrow nothing had been resolved. Things then went from bad to worse as the Air Traffic Controllers went on their major 6-month strike just as my plane was taxiing. Technically we had departed and were the last plane leaving the UK before the strike! We sat on the runway at Heathrow between 3pm and midnight arguing over whether we could take off or not. Finally the plane left, had trouble with its undercarriage and got diverted to New York for repairs. It eventually landed in Boston at 5am the next day, 19 hours for a flight to Boston! I had a breakfast meeting with Apollo at 8am. I got out of the airport at 6am and managed to get a taxi driver who had only started that morning and he got lost! I eventually arrived at the Hotel 30 minutes before the meeting with Bill Poduska. I was unimpressed by their plans. It was clearly seen as a mini-PRIME. There were no plans to put Unix on the machine, they had a proprietary LAN technology, it was keyboard driven etc. I left Boston convinced that the PERQ was still the better choice. At Three Rivers, 9 of our next batch of systems were in the shipping bay. Three Rivers were planning to fund a Unix implementation at Toronto University. CMU were developing the ACCENT kernel and SPICE operating system with Three Rivers. Tony Williams and I were both impressed by the progress that had been made. The possibilities for a good Unix implementation were there. CMU already had 30 PERQ systems in use. I left Pittsburgh on 15 August with Robb Wilmot scheduled to arrive soon after to sign the deal.
On Wednesday 19 August while visiting IBM I received a phone call from Apollo saying that they were just about to sign a deal with ICL and would SERC be prepared to switch from Three Rivers to Apollo. Robb Wilmot had visited Apollo, been impressed by their set up and been less than impressed by Three Rivers which was still in the old railway building in the middle of Pittsburgh. To aid expansion, Three Rivers wanted an up-front payment of $5M from ICL in order to expand to meet their demand while Apollo were happy to accept the terms offered by ICL.
ICL wanted SERC to switch to Apollo. Ninian Eadie visited RAL on Monday 24 August. Apollo came on the Tuesday. They offered ICL exclusive world-wide marketing outside the USA and Japan and they had plans for a $5K system. I was due to give a talk to IBM in San Jose on the Wednesday when I heard about these developments. I left Cliff Pavelin there to give the seminar and went straight to the airport to catch a plane back composing a summary of the two ways forward on the plane. At about the same time Ed Fredkin flew on Concorde to the UK to see if he could put the Three Rivers position.
Apollo really wanted ICL as a marketing arm which suited ICL in the current state, but really did not offer much in terms of the UK being a force in the workstation market, so it was less attractive to DOI or SERC. There were meetings with ICL, DOI, Apollo all day Thursday. A meeting was arranged between ICL and SERC for 5.30pm on Friday. During Friday, Wilmot met with Fredkin to hammer out a deal with ICL wanting early delivery of 40 systems in kit form and a more detailed business plan from Three Rivers.
At the meeting with ICL at 5.30pm, Rob Witty, Tony Williams and myself had put together a technical comparison of the two systems. Ninian Eadie gave the ICL position that they could work with either company. SERC's position was that the PERQ was the better product and we could buy from Three Rivers, do a deal with ICL or find another UK company like GEC to do a deal with. We would not drop the PERQ.
We left the meeting convinced that ICL would switch to Apollo. Robb Wilmot was clearly leaning in that direction. As we had 11 PERQ system stuck at Three Rivers which we had ordered from ICL, Geoff, Rob and I went to see Ed Fredkin to renegotiate the order direct with Three Rivers. We had dinner with Ed Fredkin who was also convinced that the deal with ICL had fallen through.
At 9pm, Ninian Eadie walked in, was rather surprised to see SERC having dinner with Three Rivers, and offered Ed Fredkin a contract to sign. Later he phoned Three Rivers to dispatch the first 11 systems to ICL. The PERQ Common Base Project was finally going to start over 2 years from when we started. Even so, ICL had the best product, the only competition was Apollo with an inferior system and Sun would not be launched for another 6 months.
On the 23 September 1981, ICL and SERC signed a Memorandum of Understanding with DOI's blessing. SERC agreed to buy 200 PERQs. SERC and ICL would develop the software necessary to make the workstation a success. ICL would manufacture PERQs in the UK within a year. We immediately ordered 20 more PERQs. All we needed was the software to implement the Common Base but that is another story that I will let Ken Robinson tell in his article which follows.
With these stories they always end with a where are they now? section. Well I am still with RAL as is David Duce. Tony Williams joined Microsoft and became the architect of OLE. Geoff Manning left to run the ICL start-up, AMT, that marketed the DAP workstation. Rob Witty is currently with GEC in charge of software directions. Most of the ICL people have since left the company. As for ICL they are the main distributor for Sun, particularly in the Government sector! All that remains is the dream. It is New Year's Eve. I sit at home typing this on my workstation. It cost a great deal less than the PERQ but it has an awful screen, much worse than the PERQ. I start up a window, it takes forever. I try scrolling some text with the editor. It is like walking through treacle. Oh for a proper editor that just needs a mouse! Perhaps I will go round to my son Paul who still has an A3 PERQ running and see what the future has in store for me given a decade or two!!
As 1996 breaks I get rumours of a small company in Pittsburgh that has a hand-held device with Java as the assembler language, a built in ISDN connection, a LAN-technology for the home and both voice and video I/O. I think about ringing ICL (but they are now owned by Fujitsu), Rob is in a powerful position at GEC and Tony is at Microsoft.... No I am too old, I will walk round to Paul's. Back to the future!
(This note is a brief and partial [in the sense of incomplete] history of the early part of the Common Base Programme (CBP). A full history would occupy a large volume, and would probably only be read by the author...)
In the mid-70's, SERC's Engineering Board asked Professor Rosenbrock of UMIST to recommend how best the Board could support provision of interactive graphics resources to grant-holders. It had been receiving many requests for a wide variety of computing equipment, primarily to meet the needs of interactive design, and was concerned over the duplication of effort into support of many platforms and development of similar software for them, and also the difficulty of transferring results of research within the research community and into industry. The resulting Rosenbrock Report recommended the setting up of the Interactive Computing Facility, a network of multi-user interactive machines, which belongs to another story; however it did note the needs for: central operating system (OS) support; standard software (e.g. for graphics); and an effective communications infrastructure via WAN provision.
The advent of powerful workstations was also foreseen; in the report these were called Single User Machines (SUMs), and were forecast to be powerful, cheap, and widely available; these factors would serve to accentuate the problems of multiple platforms.
(At the time, and for a long time afterwards, PCs were extremely limited in power, memory, graphics capabilities.)
In the event, the workstation scene arrived earlier than expected; the PERQ arrived in the early 80s, offering 1024 × 768 screen resolution on an A4 screen, a mouse as input device as well as the keyboard, high power graphics, and window management, all in a proprietary OS. The machine was soft in the sense that the instruction set was microcoded, and could therefore be extended or replaced. The CPU was capable of an impressive (for the time) 750K operations per second. The hardware graphics support for raster operations could only be described as staggeringly good, and is impressive even by today's standards. (I remember, as secretary to a SERC committee at the time, attending my first demo of the PERQ; as secretary, I was at the back, so stood on a chair to get a better view. It was difficult to avoid falling off; the contrast to the then norm, Tektronix storage tubes connected via a 1200 or 4800 Kbps line to a shared machine of about 250Kips, was so great!)
Only one other device, the Apollo, arrived at similar (but later) time. The first version was much less impressive than the PERQ; its graphics performance was much poorer. It also had a proprietary OS. There was definite scope for really powerful, highly interactive machines, but problems with a variety of OSs, graphics systems, window management systems, etc. would be major. There was an obvious need for standard platforms for many grant holders to aid in transferring results, and preventing research time being wasted on basic systems features. A cooperative venture between the PERQ manufacturers, ICL, and SERC to build a basic delivery system was agreed, featuring software that would be long-lived in terms of applicability. The agreement covered development of a standard OS, window management, graphics packages, and software development tools - the Common Base.
Although the Common Base was seen as a hardware initiative, in fact the objective was for it to be software dominated, particularly in the longer term. This would ease the problems of transfer of research results, and influence the suppliers to providing platforms that conformed with the Common Base, broadening the eventual choice of suitable hardware. The Unix OS was seen as the key here. The initial Common Base included:
In the early years, the Common Base Programme negotiated bulk deals with suppliers, so that researchers could take advantage of increased discounts.
Software was supplied and supported centrally.
| Category | Choice | Comment |
|---|---|---|
| Hardware | ICL/3RCC PERQ | At the time, the right choice; but the OS and hardware developments took too long to come to fruition, and finally PERQ lost its place in the workstation market. |
| OS | Unix | Correct. Even a strong market player like Apollo was forced to provide Unix as the OS eventually, despite initially commanding market presence. |
| Compilers | FORTRAN, Pascal | Correct |
| Graphics | GKS | Correct, became ISO standard. |
| LAN | Cambridge Ring | Wrong, despite technical superiority over Ethernet. |
| WAN | X25 Coloured Books | Correct - the de facto academic standard for the 80s. |
| UI Toolkit | WW | Correct - this predated a viable X Window System by some years. |
The major factors were the hardware developments, and the OS developments. On the hardware front, there were many plans for new systems from Three Rivers Computing Corporation (3RCC) - faster machines, larger memory, colour displays, etc. In practice few of these actually came to fruition. ICL's initial approach was to re-engineer the PERQ1 to make a more manufacturable and robust system, which they succeeded in doing; on the way larger discs, larger memory, more microcode space, and a 1280 × 1024 A3 screen arrived. Development was also undertaken to prototype of the PERQ3, which exploited the now powerful and rapidly evolving Motorola 68000 chip series, in parallel with the fast raster operation support that gave the PERQ its characteristic high quality graphics. This was an impressive beast, but ICL decided not to bring it to market.
Two strategies for developing Unix on the PERQ were undertaken - one of these, led by RAL, was to use the ACCENT kernel of the SPICE project under development by Carnegie Mellon University (CMU). This offered a full 32 bit paged virtual address space, in a distributed environment. (The SPICE project eventually resulted in the Mach distributed OS, some years later.)
Although this was an attractive route, as it gave many advanced facilities and excellent growth potential for the future, in practice performance was a major issue. Tests showed that there was no single bottleneck, and the development had to be abandoned as a result.
The other approach, led by ICL, was to replace the microcode by an instruction set appropriate to a Unix implementation - this was called PNX. This had the advantage that all the development was under ICL control, and the microcode could be tuned to give the desired performance. The major drawback was that, given the space available for microcode in the PERQ, a virtual memory system was not possible.
Initially both strands were followed in parallel, but ICL decided (against RAL advice) to stop development of the microcode version as the ACCENT-based port was going well. In the event, the PNX development had to be resuscitated at short notice, and the RAL team switched to porting Unix to the PNX kernel. This all took time, when there was intense pressure to deliver systems to the community who were waiting for them to do research!
Later, other tools were added to the Common Base, including the NAG Library. Following an evaluation of new workstations in 1984, the Sun Microsystems range was added. This (of course) had yet another proprietary window management system, and a Graphical User Interface (GUI) toolkit called WW was developed offering vendor-independent support for interactive graphical operations.
The impact on the UK academic community and UK industry is difficult to estimate. It is certain that many researchers were impacted by the various delays in the PERQ programme; others less so. It is equally true that the initiative brought workstation capabilities to the notice of the community probably a year or two earlier than would have happened otherwise. On the whole, Sun Microsystems probably benefited most from the Programme, as there was in place a community requiring their devices looking for a more modern replacement for the PERQ.
UK industry, in the shape of ICL, probably had most to gain. They had the opportunity to be a major player in the workstation market, and it would be easy to say they let the opportunity slip. In practice, I suspect the level of investment necessary to be a serious international manufacturer in this area may well have been beyond them at the time.
On a personal note, on taking over the Programme in June 1982, I received an interesting introduction to the management of a large national IT programme and the workstation scene generally. I look back on the episode with mixed feelings: regret for opportunities missed, a sense of achievement from the things that did work well, and getting to know a large part of the community over the years.
In this section I have tried to cover the majority of the other aspects of the Interactive Computing Facility which are not covered by the other articles. In many cases they are my recollections and my personal views, so I apologise in advance for any errors either in the facts or the dates and any omissions.
The Interactive Computing Facilities Committee (ICFC) was set up in early 1977 by the Engineering Board (EB) of the then Science Research Council (SRC). Its first responsibility was to advise the Engineering Board on the implementation of the overall programme contained in the Rosenbrock Report. From its first meeting on 28 April 1977 through to its final meeting on 27 August 1980, ICFC maintained a strong and consistent commitment to the defined programme and, through its two chairmen, Professor Rosenbrock until May 1978 and Professor MacFarlane from then to the end, fought successfully to preserve the funding, or at least the majority of it, required to deliver the programme despite numerous financial crises during the period. With such a strong Committee behind it the ICF had the best possible start and much of the credit for its success over the 14 years or so it existed must go to ICFC.
Inevitably when things are going so well something happens to change it, usually for the worse. In December 1980 the report of the SRC Computer Review Working Party (CRWP) was delivered to the Council. One of its main recommendations was that Interactive computing should be extended to a wider cross section of the SRC community. This will require the integration of the Interactive Computing Facility and its associated funding with the central resources of the Facility Committee for Computing (FCC). It would also benefit from the improvement of the front-end facilities at both main sites (i.e. Rutherford Laboratory and the Daresbury Laboratory). This recommendation was accepted, along with all the others, and responsibility for the ICF, apart from the Applications Programme (this inc1uded the Special Interest Groups) which remained as an EB responsibility, was formally transferred to the Central Computing Committee (CCC) on 1 September 1981. CCC was the new committee set up by SRC to replace FCC following the acceptance of the CRWP report. EB decided to set up its own computing committee, Engineering Board Computing Committee (EBCC) to take responsibility for all aspects of engineering computing. The Applications Programme came under EBCC and as such remained fairly well protected from the difficulties faced by the rest of the ICF that came under CCC control. CCC also took over responsibility for the Single User System (SUS) Programme.
The next 3 years proved to be more difficult for the ICF under CCC and many, but not all, of the plans set in place by ICFC before its demise had to be either delayed or abandoned. Yet another Computing Review Working Party (CRWP) was set up by CCC 1983/4 to help it prepare for its next Five Year Forward Look. One of the main recommendations of this CRWP was (surprise! surprise!) that responsibility for the ICF, and incidentally the SUS Programme, should be returned to the EB. Council approved virtually all the recommendations of this CRWP by December 1984. Unfortunately by this time EBCC had been demoted to a Sub Committee (EBCSC) which reported to the Information Engineering Committee (IEC). A further complication was that IEC had another Sub Committee, the Computer Science Sub Committee (CSSC), and it was decided that the two Sub Committees should have the same chairman and some common members!! The EB took formal control of ICF from 1 April 1985, with EBCSC responsible for the ICF.
In anticipation of the outcome of the CRWP recommendations, EB took a flyer in September 1984 and decided that, since it was almost 10 years since the Rosenbrock Report was published, it was time set up a new Working Party to conduct a strategic review of Engineering Board Computing. This was chaired by Professor Bush, who was a member of EBCSC, and was asked to report in early Autumn 1985. The somewhat ludicrous committee/sub committee structure lasted until September 1985, at which point EB decided to establish the Computing Facilities Committee (CFC) to take responsibility for engineering computing and reporting directly to the Board. Common sense had prevailed at last!!
Fortunately the much delayed Bush Report had no serious effect on CFC and it remained as the committee responsible for EB computing until it was eventually disbanded in mid-1991.
Below are my personal recollections of the birth of the ICF. As this was over twenty years ago some of the events are vague (and my memory is not what it was), so please accept my apologies for any erroneous information.
In 1975 the Rosenbrock Report recommended that the SRC (as it was then) set up the Interactive Computing Facility (the ICF, later to become the ECF). At that time most of the computing was batch based, with remote job entry (RJE) facilities providing access for remote users. Interactive computing was beginning to gain favour; several large SRC funded projects had been provided with leading edge interactive computers and the rationale was to extend this type of service to the wider SRC community. Two major SRC provided interactive systems were the DEC System10s at Edinburgh and UMIST; these supported Artificial Intelligence (AI) and Control Engineering design projects, respectively. The proposal was to upgrade these systems, which had already developed beyond their initial objectives and were providing facilities for their local communities. Additionally, it was proposed to install the latest and most powerful DEC System KL10 at Rutherford Laboratory, which was announced but not yet available, to provide a wide geographic cover and sufficient power to meet the anticipated demand. In the event the purchase of the KL10 did not go ahead, otherwise the story of the GEC and Prime Systems would no doubt have been very different.
It is worth noting that the upgraded DEC System KI10s at Edinburgh and UMIST had approximately one fifteenth of the CPU power of a 486PC, 128 Kwords (32 bit) of memory, equivalent to 0.5MB, and around 600MB of disc capacity (2 × 300MB removable disks).
Several major challenges faced us; to build a network to provide interactive services to SRC funded projects from various Universities and Polytechnics around the country; to build a support service to provide user services to these projects; and to develop a system to account usage in a simply way, thereby allowing the various grant awarding committees to easily allocate resources to projects and to know the cost of these resources.
The network was developed between the DEC System 10s, using DECNET (a set of X25 like communications protocols), and the various Universities and Polytechnics were connected via a 9.6K bit leased line network, using asynchronous multiplexors to support up to four Tektronix 4100 series graphic displays at each site. (This was state of the art networking at the time).
Initially a lot of site visits were needed to set up the network and to provide initial training and support for the remote users. Good working relationships were developed with the local computer centres as they tended to provide a degree of first line support for the project holders. Ian Cook was heavily involved in setting up the user services and was more like a travelling salesman in the early days. Ian is now Head of User Services at Manchester Metropolitan University.
The well known Allocation Unit (AU) was developed to account for resource usage and this was implemented as the method of accounting on SRC systems for many years.
In the early days the DEC 10s would crash frequently and on some occasions would be out of service for a few days (we had an on-site engineer as part of the maintenance contract).
Despite all the difficulties many leading edge applications were developed on the systems, some of which are still in active use today. (They have of course continued to be developed and don't resemble their early ancestors). Two of these packages spring to mind, these being:
I remember the Lift Systems Design Suite well as I was heavily involved in the development of this and did most of the conversion work to move it to the Prime systems and then to its current home, a PC. This package was the brainchild of Dr George Barney, who is well known to many academics and once tried to sell the UMIST Control Engineering Department, via an advert in the national press, during the government cuts in the early 80s.
The DEC 10 at UMIST survived until 1981, when it was replaced with a Prime 750, as part of the Multi-User Mini (MUM) programme.
Significant contributions to the service were made by many other people and I would like to mention a few of the UMIST staff who are still in the Manchester area (apologies to those not mentioned). Ian Pallfreeman was the systems programmer for a significant part of the time and developed many of the services such as the archiving and backup systems. Peter Mills, who latterly ran the system and communications group, Ken Roberts who was heavily involved in setting up and maintaining the leased line network throughout the whole life of the service and my secretary, Clair Sacks, who organised us so well over the years. The one time Mike Jane did not allow her to book his hotel for an overnight stay in Manchester he finished up staying at a hotel of ill-repute, but Mike can tell that story.
Those were halcyon days when computing was exciting, funding (at least by today's standards), was not too difficult to obtain, and many lasting friendships were formed which have long outlived the services themselves.
The 1980s was the decade when the SERC Primes waxed and waned. The SERC rolled out Primes as part of the MUM (Multi-User Minicomputer) Programme. At one point there were 22 SERC Primes. Like the GECs, the majority of the Primes were distributed to university engineering departments around the country. Together with the GECs. the Primes were networked to one another and RAL to form SERCNET, which was a forerunner of JANET. The rest, as they say, is history.
In the early eighties, distributed computing was in its infancy. New methods had to be devised to support grant-holders and their systems. In 1982 I joined the SERC Support team at UMIST, where we provided user support for all Prime users, while system support was provided jointly by RAL and UMIST. By 1982 much of the support infrastructure was in place. Besides the physical links between the systems, my new colleagues had been developing coloured-book software to deliver network services such as FTP (blue book) and remote login connections (yellow book). Over time, the UMIST team developed software to distribute and update the software on the remote Primes and a file archiving system whose functionality exceeded anything Prime themselves produced. We also developed an email package, initially providing a similar user interface to that on the SERC GECs.
The Prime's operating system, PRIMOS, was one of the last propriety operating systems supported by the SERC. In the early days there were annual revisions of PRIMOS, although thankfully this rate slowed down, with 1983 seeing Revision 19.1 being released. From then on there were point releases every 15 months or so. The systems team had to extensively modify the early releases of PRIMOS to provide support for networking, a batch and a spooling system.
Users were encouraged to submit enquiries, unsuitably classified as Gripes, through an email system which by to-day's standards would be considered basic. An extensive on-line Help facility was developed, which covered both the SERC added and Prime commands and libraries. The UMIST team backed this up with an extensive range of documentation: Reference and Applications manuals; a Primer; an Operations manual; a Newsletter and syntax cards. Regular training courses were held at UMIST, RAL and occasionally other sites.
Our geographic spread from UMIST to Sussex and Warwick to East Anglia led to us complementing our network support with twice yearly site visits, quarterly Prime Managers Meetings and regular SERC Prime User Groups. All these occasions were well attended, reduced isolation and allowed formal and semi-formal interaction between the support teams and the user base.
The development of the Primes was a team effort coordinated from RAL. Whenever new hardware was added to a Prime, displaced hardware would be reallocated to another system. I am sure that I'm not using rose tinted glasses, when I reflect upon the co-operation given by colleagues and users alike. Nor how prior to our User Meetings we would lunch together, sort out a series of niggling problems, so that by the time we'd reached the chocolate mousse there were few problems to discuss in the meeting proper. (The Editor's brief was to write about the SERC Prime era, and to work in a reference to the chocolate mousse that was invariably served at a London college - almost certainly the location of the European chocolate mousse mountain at that time!!). Ah those were Prime times.
The Rosenbrock Report had raised the issue of upgrading existing university mini-computers to multi-user capacity, envisaging that up to 5 machines should become part of the Interactive Computing Facility (ICF). In all, 7 machines were eventually approved for upgrades between March 1977 and October 1978. These were located at:
Apart from Leeds, the ICF provided the capital for the upgrades to each machine, additional terminals and contributed to the recurrent costs of running the machine. In the case of Leeds, the University Grants Committee provided the capital on the understanding that the ICF would provide the recurrent support.
An agreed percentage of the resource of each of these machines was made available to SERC. principally Engineering Board, grant holders with these resources managed in the same way as for the standard ICF machines. Attempts were made to define a resource Allocation Unit (AU) for each of the upgraded machines in line with the AU definitions for the standard machines, with varying degrees of success.
The unfortunate thing about these machines was that they were so different in both type and nature that there were no argument to hold regular meetings of the users or managers of these facilities. In fact as I recall there was just one meeting of the managers on 4June 1981.
What is certain is that each of these upgraded facilities provided a valuable additional resource for the ICF and the grant community. However in terms of management they were not as straightforward as the standard machines.
One important component of the Interactive Computing Facility (ICF) identified in the Rosenbrock Report was the need to provide appropriate terminals for the users. The terminal types suggested ranged from simple, inexpensive alphanumeric devices to the more expensive graphics devices (from the Tektronix 4010 to the refresh systems for dynamic displays with a light pen or equivalent). He recommended a sharing level of 5 users per terminal. Requirements covered the needs of the multi-user machines as well as that of individual grants. A Loan Pool was deemed to be the preferred mechanism since it offered the possibility of recycling terminals as some grants expired and new ones started. A set of standard devices was to be defined and refined as time went by. To ensure optimum availability a centrally managed, UK-wide maintenance contract was envisaged for all terminals in the Pool.
The first batch of terminals, costing around £100K, was purchased during 1976. This provided 20 simple VDUs, 23 basic storage graphics devices (T4006 and T4010), 7 expensive storage graphics devices (T4014) and 6 Tektronix Hard Copy Units. Immediately after the Council approved this purchase it put a stop on any further major capital purchases in the 1976/77 financial year (yes they had financial crises in those days too!!), thereby delaying the planned expansion of the terminal pool by several months. However ICF was able to place an order for its first IMLAC PDS4 Refresh Display in January 1997 for delivery in July.
There followed a steady increase in the size of the terminal pool over the next 4 years at a typical capital cost of £200K to £250K per year. New standard devices were added as they became available. One example of this was the Sigma Graphics Option Controller (GOC), which became available in 1978/79 and began to offer a politically welcomed UK supplied alternative to the Tektronix T4010.
In October 1980 there were 430 terminals in the Pool, representing a capital cost of around £l.5M and an annual maintenance cost of £110K. With 1500 users this meant that the average of 3.5 users per terminal was significantly better than the 5 to 1 envisaged in the Rosenbrock Report. A further capital injection of £655K in the 1980/81 financial year increased the size of the Pool to 539 and the annual maintenance cost was now £190K. The management of this large set of terminals required more than one person full time. The breakdown of the Pool was as follows:-
Further purchases over the few years both increased the size of the Pool and enabled the replacement of obsolete and beyond repair terminals. In 1983 the maintenance arrangements were changed to a system of central, overnight (by overnight courier) replacement of faulty devices which were returned, ideally by the same courier, to RL. Once returned the faulty devices were either repaired on a time and materials basis and returned to the Pool or scrapped depending on the estimated repair cost. This proved to be a very cost effective move which, in hindsight, should have been introduced much earlier and would have released funds to expand the size of the Pool even more. The total number of terminals purchased for the Pool was 1141, although I am afraid the detailed records do not exist to allow me to give any breakdown of the final inventory.
No new terminals were purchased after 1984 (the very last purchase was a Cifer 2634 on 3 October 1984 which is still in use at RAL today!), when the focus for new capital had switched to the purchase of Single User System (PERQs at that time). The next 4 years to mid-1987 saw a gradual reduction in the size of the Pool as terminals were scrapped, although it was still very much operational and the hot replacement system was still in place. By October 1987 the Pool was down to 20% of its peak level and a plan was in place to transfer ownership of the remaining terminals to the current holders over the following 6 months. This was clearly a very sensible course of action. Such actions were not always associated with SERC at the time! I am sure the useful life of the terminals was greatly enhanced by this action and that users were able to continue and complete their research projects more effectively as a result.
ICF put a lot of money and effort into the provision and support of the Terminal Loan Pool over the 12 years it was operating. The combined total capital and recurrent costs excede £3M, excluding the RL based effort which was in excess of one full time person during the peak period. In terms of the Rosenbrock target of 1 terminal for every 5 users, this was often bettered and at one time was operating at 1 terminal for every 3 users. It represented a key component of the whole facility and played a significant role in the success story.
Applications software was an essential component of the ICF. The Rosenbrock Report emphasised the need to avoid wasteful duplication of effort in the development of software packages and to provide adequate documentation for general use, with provision of such packages and their support only at a single site. It also recommended that resources should be made available to finance applications workshops, conferences and development of the Meeting House function.
In September 1976 the Interactive Computing Facility Committee (ICFC) requested the then Rutherford Laboratory (RL) to organise 5 one-day meetings to assess user requirements for applications software support in selected areas. The areas were Finite Elements. Electromagnetic Design, Artificial Intelligence (AI), Circuit Analysis (Digital Systems and Integrated Circuits), and Circuit Analysis (Lumped and Continuous Circuits). The meetings reported to ICFC in March 1977 and as a result 5 informal groups were set up for one year to address these areas and advise RL about the short term needs and estimate effort needed over 5 years. The financial provision was for 10 staff-years (sy) of effort at RL in 1977/8 and 8 sy in 1978/79.
In November 1977 ICFC approved funding for 3 years to take over support for two programmers at Edinburgh who were already supporting the AI community. At the same meeting, one-day meetings were approved to discuss the needs for an applications focus in Computer Aided Architectural Design (CAAD) and Control Engineering. The requirement for ICF to establish a number of Special Interest Groups (SIGs) was discussed and it was agreed that a SIG should be supported for a maximum of 5 years.
In May 1978, the 5 informal groups reported on their needs and ICFC approved the establishment of 4 SIGs in Finite Elements (SIGFE), Electromagnetics (SIGEM), Artificial Intelligence (SIGAI) and Circuit Design (SIGCD). The latter being a result of combining the two informal groups into a single SIG. Three further SIGs in CAAD, Control Engineering and Tools for Interactive Programming (basically an infrastructure SIG to underpin the others) were established over the next few months.
Over the next several months these SIGs addressed their detailed funding requirements and successfully made their cases to ICFC for their work programme. Initial funding ranged from 3 to 5 years. By the end of 1981 all 7 SIGs were fully established and reporting regularly to ICFC on their activities, which were concentrated on developing and procuring applications packages and supporting them across and beyond the ICF. The majority but not all of the effort was based at RAL. SIGCAAD was run from Edinburgh and there appears to be little or no information available (at least to me) on its activities beyond 1981.
So what did the other 6 SIGs achieve during their lifetimes? The following summary is extracted from an article written by Alan Bryden in the Engineering Computing Newsletter of May/June 1987.
The SIG originally recommended two approaches:
It was intended that SIGFE should regroup as a Numerical Modelling SIG following an open meeting on October 1984, but because of the uncertainties following the Bush Report on Future Provision for Engineering Computing, this has not happened.
SIGEM was the longest survivor of the SIGs holding its final open meeting on 8 April 1987. Most of the software supported by SIGEM was developed at RL. The three major packages, PE2D, TOSCA and BIM2D are for electromagnetic and electrostatic analysis in 2 and 3 dimensions. The programs were heavily used by universities and industry. (Note: The modern versions of these packages are now available through Vector Fields). A sophisticated post-processor RUTHLESS, developed by Infolitica Ltd, can now be interfaced to any 2-dimensional electromagnetics package.
In addition to being the longest survivor, it is fair to say that SIGEM proved to be the most successful of all the SIGs without detracting from the success of the others.
There were originally two SIGs covering Digital and Analogue Circuits. These were merged with a parallel activity promoted by the Solid State Devices Subcommittee on design, placement and layout for digital ICs. Later, all of the programmes in IC design became the responsibility of the Microfabrication Facilities Sub-committee of the Information Engineering Committee (IEC), and there is now a management committee for the UGC/DTI/NAB/SERC ECAD Initiative, chaired by Professor Dagless (Bristol University). The remaining activities in device and process modelling are focused in two projects funded by the Alvey VLSI programme, managed through the Process and Device Modelling Working Group, and in a project funded by the EEC's ESPRIT Programme.
The SIG supported work at Kingston Polytechnic on provision of the SLICE library of subroutines, which I believe is still available through NAG. The SIG disbanded in 1985 when the Management Committee for the CDTCE (Computing and Design Techniques for Control Engineering) Initiative was set up.
SIGAI was originally formed to support and extend the languages such as POP2, LISP and PROLOG used by AI researchers on the DEC10 at Edinburgh University. The SIG has now regrouped and reports to the Alvey IKBS Director.
The SIG supported basic tools ranging from Geometric Modellers and Mesh Generators to Command Decoders and Graphical Input aids. The SIG has not been active since mid-1985 because of the priority given to the CAD*I data exchange project funded by the ESPRIT Programme."
In March 1977 two papers were presented to the Joint Edinburgh/UMIST Management Committee proposing a charging and allocation system to provide a mechanism for effective resource control for the two DEC System 10s.
The objectives of the proposed system were to:
Control was exercised at two levels, the project allocation and the period ration, where each period was defined as 4 weeks. The Resource Unit for each DEC System 10 was defined in terms of 4 units of measurement, CPU time (C) in hours, Disk space (B) in Kbyte blocks, Connect time in hours (T), and Core occupancy in Kword hours (K). The initial Resource Unit was defined as:
R (UMIST) = C/0.7 + B/120 + T/20 + K/21
R (Edinburgh) = C/0.7 + B/240 + T/25 + K/28
The different values of the constants reflected the different machine configurations.
Each Resource Unit was weighted by a time-of-day factor to turn it into an Allocation Unit (AU). The initial factors were Peak (1.0), Standard (0.5) and Discount (0.2). This led to an initial annual number of AUs of 10,000 at UMIST and 12,500 at Edinburgh.
Each AU (Bob Hopgood assures me AU was chosen because these resources were like gold dust!!) was assigned a notional money charge rates of £15 (cash cost), £22 (full cost including capital recovery), and £44 (market rate for commercial users). This enabled committees to consider the true cost of grant applications by adding the real cost of the grant application and the notional cost of the ICF resources requested. The assigned value was also intended to enable an easy transfer of resources from one interactive system to the other. This was in effect the ICF's equivalent of the ECU in Europe today. In fact I suppose it could have been decided to use ECU (Engineering Computing Unit) rather than AU at the time!! The discounted time in unsociable hours was designed to encourage users to expand the interactive day outside prime time.
In January 1979 the charging algorithms were reviewed in the light of the early experiences on the two DEC System 10s and a revised system was approved which covered all ICF machines, adding in the GECs and Primes. The Resource Unit for each machine type was defined and the time-of-day factors modified to Peak (1.0), Standard (0.4), and Discount (0.1). The full cost rate per AU was set at £25 across all machine types.
The ICF management took the AU very seriously and a lot of effort was deployed in the allocation and management of the allocations at the individual grant level. I am sure that every recipient of a grant to use the ICF and every MUM Manager will recall the infamous Pinks issued by the Resource Management staff!! At the peak there were 3 people working full time on the resource management activities. Grant Holders also took the allocations very seriously, to the extent that they took full advantage of the unsociable hours discounts to extract the maximum possible from their allocations. The heavier loaded machines such as the DEC System 10s had a significant level of interactive usage at the discount rate (i.e. users were active throughout the nights and weekends).
Over the years there was a lot of lively debate about the philosophy as well as the actual implementation of the AU within the community. On balance I believe the system, however imprecise, did work effectively and provided as fair a system as any other in terms of allocating and delivering interactive computing resources to the large user community.
The ICF newsletter, Rapid Response, was first published in February 1977. Its purpose was to inform existing and prospective users about the ICF and was another implementation of a Rosenbrock Report recommendation. The publication frequency oscillated between every 2 months and every 6 months until the final issue (number 13!!) which was published in November 1980.
The transfer of responsibility for the ICF to the Central Computing Committee at the end of 1980/81 meant that from this point until the end of 1986, almost two years after responsibility for the ICF was returned to the Engineering Board, the scope of the existing Computing Department newsletter, FORUM, was extended to include information on the ICF as well. Although providing a reasonable mechanism for imparting news to the ICF community, the much wider remit for this publication meant that the ICF information was embedded in other information on the mainframe service, etc. and as such lacked the real focus which Rosenbrock intended.
Eventually in January 1987 the Engineering Computing Facility (ECF), the result of combining the ICF and the Single User Systems activities, published the first issue of its new newsletter, Engineering Computing Newsletter. As you are no doubt well aware this publication has continued up to this point.
Everyone associated with the three publications during the past 19 years worked hard to ensure that the engineering user community we were funded to support was kept well informed on all matters, both technical and political. We can only hope we succeeded!
During the period 1989/90 there was a series of SERC visits to universities and (then) polytechnics. Each visit lasted a day. These visits were part of the EASE programme and consisted of an introductory presentation followed by a series of interviews with engineers about their use of IT in their research.
The introductory talk had to assure the audience that the (then) SERC would not build league tables from the information collected on the visit. This promise was kept, but it was quipped that we might build and publish a league table that ranked the quality of the lunch.
Enough time has now elapsed to allow the memory to become less precise and the catering arrangements to improve. The following table is incomplete but interesting nevertheless.
| HEI | Rating | Comment |
|---|---|---|
| Aberystwyth | ***** | Can't remember the food, but the view across Cardigan Bay on a glorious summer's day was unforgettable |
| Bangor(UCNW) | **** | Red Dragon pie at bistro just off campus |
| Bradford | *** | One of a series of visits made in December when Christmas lunch was de rigueur |
| City | **** | Usual impeccable food in the restaurant, no wonder we used City as a regular venue for meetings |
| Durham | * | The morning schedule overran, so we had to scavenge in the local snack bar |
| Keele | *** | Pub grub overlooking a golf course |
| Leeds Poly | *** | Started to worry about my waistline |
| Paisley | *** | Local Chinese restaurant just behind the College |
| Salford | **** | As Bradford, but impressed by stunning paintings of Manchester United in their glory days |
| Sheffield Poly | *** | As Bradford |
| UMIST | *** | Booked the Private Dining Room myself; I wasn't disappointed |
| Wolverhampton | ** | This was our first visit before I was into the swing of organising the visits |
Unfortunately in this life all good things must come to the end and ICF was no exception to this. From the outset in 1976, when the only two machines in existence were the DEC 10s at Edinburgh and UMIST, through to 31 March 1983 the ICF grew significantly. From then on there was a mixture of enhancements (the last major machine purchase was a P9955 to replace the two P750s at RAL in early 1986) and machine closures right through to the very end when Professor MacFarlane formally switched off the last ICF machine, the Prime 9955 (RLPA) at the Rutherford Appleton Laboratory, on 30 March 1990. The Engineering Board agreed that all machines where support contracts were terminated should be offered to the local site. In many cases this offer was taken up and local users benefited from continued availability of these machines.
The first signs of reduction came at the end of March 1983 when recurrent support for the upgraded machines at Sheffield and Swansea ceased. By the end of 1984 support for the upgraded machines at Leeds, Oxford, Manchester Graphics Unit and Southampton had terminated. The upgraded Primes at University of East Anglia and Nottingham had been adopted as standard ICF machines in 1981.
1985 heralded the start of a major rundown programme. The heavily ICF modified version of the GEC OS4000 operating system was frozen. The Nottingham Prime and 6 GEC machines (at Bradford, Cranfield, Glasgow, Heriot-Watt, Newcastle and Queen Mary College) had their contracts terminated on 31 March. The most powerful ICF machine, the DEC 10KL at Edinburgh, was formally switched off by Bernard Loach on 30 September. The Edinburgh machine had served SERC grant holders for nearly 9 years, providing a highly professional service to a very large user community. (Incidentally the machine was moved to Essex University and was installed alongside their existing DEC 10).
In the period March 1986 to March 1988 all remaining remote machines had their support terminated. The SERC version of the Prime operating system PRIMOS was frozen in 1987. The last GEC service machine, the RAL 4090, was switched off in December 1988.
The formal end of the ICF was reached on 30 March 1990 when Professor MacFarlane, who had played a major role in the successful implementation of the Rosenbrock Report recommendations, switched off the last ICF service machine, the RAL P9955, at a ceremony at RAL attended by around 50 invited guests.
During its 14 year history (1976 - 1990) ICF proved to be one of the Engineering Board's major success stories. A total of around 20,000 users benefited from the wide range of services offered on the various machines. The estimated total cost of the ICF service, including capital, maintenance and staffpower, was £20M.
It is difficult to estimate the total number of people who were involved in the ICF in one way or another during its 14 years existence. My own personal recollections of the people are all very positive ones.
The usual interaction of people was at (and after!!) meetings of one sort or another. There were regular meetings between the then Rutherford Laboratory (RL) staff and users (DEC 10, GEC and Prime), managers (GEC and Prime) and management (contract reconciliations/negotiations) as well as other meetings with individuals for various reasons. These meetings enabled people to get to know each other and gave a sense of belonging to something important to which they were able to make significant contributions at the individual and collective level.
I recall long train journeys (e.g. to Edinburgh when all planes were grounded and regular trips to Manchester on the 0702 from Oxford - surely BR's worst train service!!) which brought RL staff together for long periods and helped them to form close bonds away from the office. It would be wrong of me to mention names since it would be impossible (even with my memory!!) to remember everybody. Therefore I will end this brief article by thanking everyone who was involved in the ICF for their dedication, professionalism, and, most important, their friendship. I personally will always look back on the ICF days as the happiest of my working life at RAL.
I will try to keep this brief since the EASE Programme should still be relatively fresh in readers' minds.
EASE was first mentioned in ECN issue 4 (July/October 1987). It was the result of the Computing Facilities Technical Advisory Group (CFTAG) review of the Common Base Policy. The EASE Strategy and detailed Workplan (for 1989/90 and 1990/91) were approved by the Computing Facilities Committee (CFC) on 4 April 1989. The EASE mission statement was To stimulate and encourage engineers to use appropriate state-of-the-art software and hardware to enhance their research. A special (undated!') issue of ECN in July 1989 informed readers about the details of the EASE programme and encouraged them to both comment on and participate actively in the programme. The strategic objectives were to:
The initial Programme was based on three major planks:
The Engineering Computing Newsletter was to be the major method of communication.
I list many, but certainly not all, of the EASE achievements below, but there was one critical, fundamental difference between EASE and its predecessor programmes in engineering computing. Unlike the Interactive Computing Facility and the Single User System Programme, EASE was not identifiable in terms of blue (GEC), orange (Prime), or fawn and brown (PERQ) boxes. It was consequently a significant (in terms of cost in excess of £3M per year) infrastructure activity that could be cut at very short notice if a financial crises was anticipated. You probably do not need reminding that this is exactly what happened in December 1991 when the then SERC believed it had a serious financial crisis in the current year and that the situation would not improve for the next two years. Its reaction was to require all the Boards to cut their programmes to accommodate this problem. In the case of Engineering Board (or more correctly the Engineering Division at Swindon Office) it was decided that the best way of achieving its share of the cuts to be imposed by SERC was to reduce the EASE Programme by 80% within one year (this meant, for example shedding 32 of the 40 staff working on EASE in just ]2 months!!). In the event there was no crisis but despite this, on the advice of its Engineering Research Commission (ERC), Engineering Board decided it would not change the decision to cut the EASE programme. To demonstrate how difficult it is to survive once the knife has been stuck into you, I remember attending a 2-day meeting of ERC in early 1992 where, on the first day, it was reported that the anticipated financial crisis did not in fact exist and therefore the Boards did not need to cut their programmes after all. Nevertheless later that day the meeting launched into its attack on the EASE programme. When I arrived at the start of the second day, I found, on the table in front of me a table produced by the ERC secretariat proposing how the money saved by cutting the EASE programme should be shared out amongst the many ERC committees!! If ever there was a case of here is something we prepared earlier this table was it. This effectively was the beginning of the end for the EASE programme. From 1 April 1993 we have seen a steady decline in both the funding and committee support for EASE. As you are well aware we have now reached the very end and EASE will cease to exist after 31 March.
I want to conclude on a more positive note by listing some of the EASE successes with due apologies for any omissions which are certainly not deliberate. To avoid, or at least minimise, any personal bias I simply list them in no particular order and without comments:
