In the early days of university computing each university had its computer centre with its computer installed as a result of the Flowers Report report that proposed various computers in various sites. These machines provided a batch service but the resources were limited in total power and the software provided.
So, university engineering departments were dissatisfied and wanted their own interactive facilities. This resulted in Professor Howard Rosenbrock of Manchester University writing a report that proposed in 1975 that engineers should be provided with interactive facilities. This was to be achieved by:-
The report was very far-sighted as it predicted the day when each engineer would have their own computer on their desk. It also had a belief that networking would be important but at that time they were very unsure of what they meant by networking. The impression was that the connections were for interactive connections and batch access to large machines such as the IBM at Rutherford.
As an aside, the only aspect of the report not implemented was the large interactive machine at Rutherford.
The project was named the Interactive Computing Facility and was led by Peter Davey seconded from Oxford University.
The report was accepted and the job of selecting suitable mini computers was allocated to Rutherford Laboratory that had just absorbed the Atlas facility. Bob Hopgood led the job that came at a propitious moment as the ICL 1906A was being phased out thus releasing some highly qualified staff.
Rather than just select the ubiquitous PDP11 the team drew up a list of 28 possible machines. This was whittled down to 10 by a paper study as many were non starters being too small, lacking suitable software or other eliminating features. The team set about evaluating them.
It was expected that FORTRAN would be the predominant language so the evaluation was based on a script that included editing, compiling and running a program. At that time there was no technology for automatically running such scripts particularly across a range of machines. Thus the evaluation consisted of 6 people at 6 Teletypes going through the script. These were known as the British Standard Idiots as mistakes were expected and what happened then was of interest.
The final set of machines benchmarked were:-
Of course, other aspects such as technology, price, size and availability were also part of the evaluation.
The exercise was highly entertaining, revealing as well as embarrassing to some manufacturers and surprising to others.
In the case of Interdata 8/32 the 6 British Standard Idiots sat down and started. As they moved to compilation they all waited and waited and eventually gave up as nothing appeared to be happening.
The surprise machine was the GEC 4080. The machine was built for real time applications but GEC had recently developed a comprehensive multi access operating system -OS4000. The machine was (and in my view still is) unique in having a built in process scheduler that makes process swapping very fast. This is helped by a built in interprocess message passing system. The machine also exhibited many aspects that we take for granted today but were novel then. These include segmentation, read only code segments, good store protection. The machine had the advantage of being British.
The other leader was the PR1ME. This was built by a new USA company (existing from 1972 until 1992) by staff from Honeywell. The machine was a clone of the Honeywell 316 and 516 machines. It was also unusual as its operating system was written in FORTRAN. The idea was that as the company developed new computers the software would be easy to port.
The Atlas Lab already had a 4080 that was being developed as a communications front end to the 1906A so that it could be networked with the IBM computers as Rutherford and Daresbury. It has to be said that the technical experts were much impressed by the machine. The X25 code developed for the front end could be easily be used in the multi user mini role.
The PR1ME also had an advantage as it had X25. However both computers had no high level protocols.
The 4080 team was led by Paul Bryant aided by Dave Toll, Philip Gladstone, Jonathan Mills and later by Leslie French and Shirley Wood, Jacky Hutchins provided user support.
The PR1ME team was led by Cliff Pavelin aided by Len Ford, Graham Robinson and Phil Newton.
PR1ME computers were installed at Rutherford and universities as follows:
Rutherford 1977 Nottingham 1977 Sussex 1979 East Anglia 1979 City 1980 Surrey 1980 UCL 1980 UMIST 1980 Warwick 1980
The Joint Networking Team (now UKERNA) had been busy defining the network protocols for use in the academic community. These were the so called Coloured Book Protocols. At the time the multi user project started very little implementation of the Coloured Books had taken place. In fact many of the protocols were still in the process of being drafted.
To leap ahead, at the time of the Glasgow Networkshop, the GEC 4080 work had progressed to the point where a file transfer protocol was needed and only a draft was available. The author was concerned that the drafting of the Blue Book File Transfer Protocol looked like it would go on for ever and said that if Blue Book was not finalised in a few months then he would implement it as is and as he was likely to install many 4080 and PR1ME computers this would prejudice adversely the aims of the protocol work. The threat worked.
The community was confident that networking was going to be based on X25 and the Coloured books with the hope that the world would follow. History shows we were wrong.
As noted, the Rosenbrock committee did not really know what they expected from networking and were probably thinking in terms of terminal access and remote job entry to large computers such as the IBM at Rutherford. The 4080 group knew exactly what they wanted to do and this was novel, far-sighted and risky.
The philosophy was to have exactly the same operating system on every 4080 and to undertake all development at Rutherford. This was expected to remove the need for any expertise in system development at the sites. For convenience this needed the ability to file transfer a new system to a remote machine. Interactive use was then needed to access the machine and reboot it with the new system. This was done late at night so as not to impact the service. This was a highly risky idea but in practice it worked well. These days the remote support of computers is well established.
E-mail was implemented as that was wanted as the main means of receiving fault reports, requests for many things. It very rapidly became popular for all sorts of things.
The team, or rather Leslie French, implemented the Red Book Job Transfer Protocol. This never became popular as running jobs on other 4080s was not a sensible thing to do. However, running jobs on the IBM machine was popular but this was achieved with NJE as Red Book was not provided on the IBM until late in the life of the project.
Rutherford developed an X25 exchange based on the 4080 written by Andrew Dann. It will be no surprise that the X25 code for the Multi User Mini came from the same source.
GEC did not expect to sell many 4080s and their production line was somewhat slow. Deliveries were always a problem. However, we understand that the salesman's commission was quite good and as a result of the project and allied sales the salesman, Peter Lane, was alleged to earn more in one year than the GEC managing director. In those days government money could not be carried over the year end and the author remembers visiting the GEC factory at Boreham Wood to examine some chalk marks on the floor and confirm it was a computer so that it could be paid for.
The first machine was deliver to Atlas building in Spring 1977 As the team had already been working with the front end 4080 the work to develop the network code was well advanced and was ready for the installation of the first 4080 in a university.
GEC computers computers were installed at Rutherford and universities as follows:
Rutherford 1978 Bristol 1978 Cambridge 1978 Glasgow 1979 Cranfield 1979 Newcastle 1979 Cardiff 1980 Birmingham 1980 Bradford 1980 QMC 1981 Sheffield 1981 Heriot-Watt 1981
Seven other GEC systems (Appleton, UCL, Manchester, Rutherford, ROE, NERC(2)), although not purchased by ICF, were linked to and supported by ICF.
Each site employed a site manager whose job was to make sure the computer was working properly, to help the users and report problems and faults.
A comprehensive manual was written that incorporated how to use the network features. This really aimed at compiling and running Fortran. However, the uses to which the computers were put was never investigated or recorded.
Manger's meetings were held once or twice a year. On one occasion a meeting was held in Glasgow and it was calculated that with the 10 or so people from Rutherford it was cheaper to hire an aircraft to go from Oxford. After the event management were not impressed and were concerned at the impression to would give even though it was cheaper.
The terminals were driven by a board with a Motorola 6800 microprocessor. This gave an easy opportunity to build other interfaces. Two were built.
The first was a Cambridge Ring interface. The Cambridge Ring was the great British local area network technology. The Atlas building, in expectation, installed a ring whose main purpose was to interconnect some early word processors. The 4080 interface was the only route to the outside world although the word processors lacked many protocols so was of limited use. The technology was not very successful due to the difficulty of building the Ring interface and lack of interest from the industry.
The second was an Ethernet interface. Interlan marketed an Ethernet board that was quite easy to interface with the 4800. Although the interface worked there were few others machines to talk to and none with higher level protocols.
GEC provided a board for high speed interfaces and a box was built to interface to the Hyperchannel network that was thought to be a local area network way forward. This was a difficult network to install as it required large coaxial cables that were difficult to thread through the ducts. The interface was never used in anger as the Hyperchannel did not develop as a way of connecting large numbers of machines.
In retrospect the teams, or at least Paul Bryant, were keen to see a local area network around the Rutherford site but perhaps ahead of their time.
The combined network of 4080s, PR1MEs and other SERC computers was extensive with links to the majority of UK universities. On the other hand, the Joint Network Team was making heavy weather of providing a national academic network. There were a number of regional networks such as Metronet in London, Midnet and South West Regional Network that were all based on ad hoc or propitiatory systems. There seemed to be considerable reluctance to fund the JNT with staff and equipment to do the job. Eventually the lifetime of the ICF started to end and SERC made the useful decision to gift SERC net to the JNT as the basis for a national network JANET that arrived in 1984. This conveniently got rid of the cost of running the network and kick started a national network. It is probably true that the ICF gave the UK a year or two start on networking with respect to other countries although the protocol wars were about to erupt - but that's another story.
