Forgotten Voice

In early 1958 I applied for a job with Ferranti Computers at Bracknell. I was unsuccessful but every cloud has a silver lining, as I will explain. I was then working on missile test gear development in Feltham and I wanted to move. Transistors were making big news and switching circuits were being developed using them. At Ferranti I used to see many faces which would later become familiar.

Later in 1958 a job advert for a company called IBM advertised for staff in the Daily Telegraph and this was intercepted by my wife! It was not a grand advert but a postage stamp square single column perhaps 4cm deep on an inside page. I fancied working for a small company so I applied and was successful.

I was interviewed at the then location of a small IBM laboratory in the UK offices in Wigmore Street. The interview was conducted by Charles Owen who tested my switching circuit knowledge which seemed to please. I was offered a job in Wigmore Street but told that the laboratory would shortly be moving to a village called Hursley. As the Lab move was imminent I opted to join in Hursley and became the first to join following the move from Wigmore Street.

Charles Owen had worked with Manchester University, Ferranti and Elliott Brothers on their ground breaking computer developments after WW2. He was for many years a senior technical contributor to most Hursley projects.

Before accepting the offer we paid a visit to Hursley. We came to Hursley House via the Winchester entrance. Hursley House had been a major contributor to the Spitfire Fighter project during the war. The cottage gardens were littered with Spitfire canopies mostly used as garden cloches. The house had recently been vacated by Vickers and looked like it; ‘A’ block as it became seemed to have a leaky corrugated roof. What a place to work, so I accepted.

As I had opted to join in Hursley I was not part of the moving process. By the time I arrived it was populated by office essentials. The ground floor including the Wedgwood room was primarily laboratory space apart from the hall, which was reception etc. and a rudimentary typing pool. Soon a travel agent was established by the main doors. The first floor was retained for offices.

The lab manager at the time was WS Elliott a Cambridge University man who was presumably charged with the initiation and planning of the laboratory. Within a short period he returned to university life and was replaced by the mercurial John (later, Sir John) Fairclough who had spent time in the USA. His first tenure at the lab saw its transformation, primarily with respect to the development of the 360/mod40 processor.

There was aggressive recruiting and space was required. The tumbledown building known as ‘A’ block was renovated to provide space for new arrivals. The progress of the lab was very much in the hands of John Fairclough and his American boss Bob Evans who became a regular visitor at this time. It was a great compliment for the lab to be awarded the design and build of the System 360 model 40 and a big departure for IBM to award such a task outside the USA.

At that time there was no catering at the lab and lunch was often a walk down to the Kings Head Pub in Hursley for a slice of Veal & Ham pie and a beer. Later Meal Vouchers were provided, but I am not sure where we spent them.

The lab quickly had a very good image. Telephones were regularly cleaned and disinfected by Phonotas. White laboratory overalls were available with a regular laundry service. Travel to London Airport for mainly USA flights was provided by Mr Raynsford, who lived in the village, or by John *******, the proprietor of the Hursley Garage. Atlantic travel to the USA by Cunard was still available from Southampton and was initially allowed as an alternative to flying.

When I arrived, a project, the 6300 technology prototype, had already been initiated in one of the laboratory rooms overlooking the drive (A Lab). In retrospect it was a project used to build skills for what was to be a quickly growing laboratory. So much for joining a small Company!

To microprogram, new term at the time, the machine would use arrays of small ferrite toroids that were currently in use for memory data storage. These toroids were mounted on an array of square moulded plastic mats and employed the hysteresis effect to generate output from a wire threaded on one axis of the array, when wires threaded on the other axis were driven or energised. This was the beginning of the Read Only Memory (ROM) technology which was to have far reaching effects and was the first IBM machine to use this form of ferrite technology.

The build structure for the machine was a Post Office frame. At that time memory array manufacture was a cottage industry where ladies from the village were employed to thread the wires on the output axis to a pattern to implement the 1 & 0 sequence of a program by missing or threading a toroid. As I had some experience of test gear I contributed the driving system for the arrays which needed flexible timing during development. The test gear, and the 6300 itself, was built by Harry Shaw who had been recruited from Ferranti along with David Truslove (To the best of my knowledge). Harry later became manager of the site workshops. The project manager was Douglas Rae, Big Duggie, as he was affectionately known, and Matt Taub was responsible for the design.

At that time a microsecond was quick and there was informal competition to run fast by driving the Mullard OC44 transistor (state of the art technology at the time). British made laboratory equipment did not meet the needs of the developers, particularly so with regard to oscilloscopes which seemed to have their heritage in radar. The round face screens were small and they were very poor for examining multiple traces. Synchronising (“syncing”) was a major problem. Top of the range was a large fridge size machine with a small cupboard in the upper half without an obvious purpose apart from keeping your lunchtime pasty hot.

Tektronix changed all that; their trolley mounted oscilloscopes were a joy to use with a clear rectangular screen. The first to arrive was awaited with anxious appreciation but quickly disappeared upstairs to John Fairclough’s office. However, we soon all had one. Bench power supplies were provided by Solatron. I believe they were absorbed into what eventually became Vodafone.

Things were on the move. The 6300 had done its job. The post office rack was found to be too small and would have to be repackaged. The machine was summarily dismembered. Its read-only memory design was soon to be replaced by far superior designs. In its place as a new project was the 832 ET Biller. This was to replace the 632 ET Biller, a desk based office accounting machine that used the rapidly outdating valve and relay technologies. It processed punch cards for input and output with an IBM golf ball typewriter on the desktop.

I was given the card reader to design and Phil Austin was given the card punch. Documentation support was provided by Bob Clayton and three of his colleagues. At that time they were a group of robust contractors who travelled to work fifty plus miles daily in a Mini. Bob eventually transferred permanently to Hursley. The project manager was Harry Metcalfe who had interviewed me at Ferranti!

The 832 used ROM technology based on a ferrite ‘U’ and ’I’ proposed by John Ayling, a key designer. It was developed at the same time as a Toroidal ROM was being developed in the Wedgwood room by a team led by Tony Proudman. These activities preceded the TROS3 (Transformer Read Only Store) developed for the S/360 Model 40.

TROS was developed from the experience gained from these two projects. The basis of TROS was a mylar tape printed with a line of connected, figure eight figures with holes in the 8. The printed pattern provided a conducting path through the tape. The tapes were stacked together with one arm of the U through the holes in the mylar and the ‘I’ then laid above to complete a magnetic circuit.

To program the memory the arm of printed 8 passing through the U and I was assigned ‘1’ and the arm outside the U and I a ‘0’. The tape was programmed by punching through one or other arm of the printed circuit ( https://en.wikipedia.org/wiki/Transformer_read-only_storage ).

The 832 in its entirety was product tested in the new Product Test facility in Millbrook where to the best of my knowledge no major problems were found. It went through a design analysis and was then terminated. The reason for this was unclear but the product contained a considerable amount of custom logic, whereas the majority of those doing the analysis were unfamiliar with free form logic. Off the shelf pre-tested logic modules, which were fully evaluated, were coming but would not have met the design needs at the time. The doubtful future of punched cards may have also been a factor, but they were still in regular use at that time. It was certainly one of the first IBM machines to pair transistors with electromechanical devices.

Its demise might have been that there was something better in the pipeline in the shape of a System/360 Model 40, development to be managed by David (DAT) Reid. The S/360 Model 40 development saw huge changes with an influx of engineers and programmers from the IBM Poughkeepsie/Endicott Labs, the mother locations for the S/360 project. New hires were common, and staff moved in from IBM UK, particularly from customer engineering. It also provided employment for many of those who had been involved in the 832 and Scamp projects.

‘A’ block was turned into a laboratory building to house the new staff. ‘B’ and ‘C’ blocks were planned and built around this time with ‘C’ block providing space for a badly needed computer centre. ‘B’ block is reputed to have been planned with two more floors than were required. As it happened the plan was passed and ‘B’ block with its relatively small footprint became known as the tallest lift shaft in Hampshire.

I became responsible for the Local Store logic design and Memory protection features and then like many others had time in Poughkeepsie during the testing period. The system was announced in May 1964.

The S/360 Model 40 was followed by the Model 44 which was a machine with extended capability for special, mainly scientific, users. It had a High Speed Multiplex channel, a High Speed Direct Data channel and Storage Protection and Memory extension. The manager for this work was Dennis Atwood, a former customer engineer. The project was managed by Alan Glover, also a former customer engineer. The S/360 Model 44 was announced in August 1965.

It was followed by a period working on the design of a high availability two-way multi-processor version of the Model 44. This resulted in a number of patents but not much else.

The next generation of mainframes (the System 370 series) was being planned where Hursley became responsible for the S/370 Model 135. Bob Jackson, a colleague from the 6300 days, returned from the USA to manage the project, after an initial period when it was managed by his boss from the USA.

There was a parting of the ways. Software and Applications were becoming increasingly important as was the importance of new technology. Systems were to become more distributed and Hursley was to have a finger in all these activities.

Hursley was developing a new programming language called PL/1 and there was work on a major software project for customer management called CICS (Customer Information Computer System) requiring lots of processing and peripheral equipment. CICS was to focus on the need for increased terminal storage.

Disk files were soon to go through a transformation through miniaturisation from large floor standing cabinets to small devices to be included in most peripheral computers. Hursley was in the forefront of this change with the Gulliver and Dolphin disk files. The Hursley disk mission was later transferred to a new factory in Havant.

The initiator of the disc project was John Herbert (Note by Editor: Might be 'John Heath') who I believe was originally an IBM sales person. Leo Rigby was a key inspiration for the technology that was developed. A key challenge was the design of the flying head for a small disc file. The flying head wrote to and read from the disk surface. Development was managed for many years by Derek Cresswell in a department run by David Cuzner.

The semiconductor was about to change the architecture of the total computer system with the advent of cheap microprocessors and semiconductor memories together with fast fibre optic communication. This was coupled with a requirement to identify new users and their technology needs.

At Hursley, it was thought that to exploit the increase in processing power that was expected to occur, new applications were needed. (How right they were!). Stimulus in this instance was provided in a number of projects including one for State Farm Insurance. It covered the needs of planning authorities and service companies to digitise large numbers of existing maps and drawings. Work to facilitate the input of large numbers of drawings into a computable form was required.

The way to convert the contents of old drawings was to scan and vectorise the content for processing. IBM produced a scanner that could handle wide documents, but software to perform a vector conversion was not available. A mainframe Pascal interpreter was available and prototype drawing conversion work was undertaken. The objective for the code was to synchronously vectorise the scanner output as it was generated. The mainframe interpretation of the Pascal was slow and limited the size of sample that could be used, however the code developed showed promise.

It worked like a knitting machine following lines using a linearity constraint developed by John Evans. It required little intermediate storage, only somewhere to put the vectors as they emerged. Characters were vectorised like everything else, as were solid areas. There was potential to cut the vector set to separate character vectors. This required that the results were passed to a character recognition process which did not then exist! An intermediate solution of an interactive workstation was a temporary, or even permanent, requirement. I am sorry to say that the appetite to undertake the required development did not exist and no product emerged from this activity. It did however focus work on more powerful intelligent terminals.

There was, to my knowledge, no attempt to focus on commercial applications for the masses even though a number of market planners were imported from IBM UK. IBM, as ever, was serving the (current then) big commercial users of mainframes.

A Flat Panel technology project and a human factors laboratory were established in ‘A’ block with an extensive investment in technology capability under the guidance of Don Barclay who was developing an electro-chromic display technology.

The termination of what had been called Industry Systems led to an activity that was to be a turning point for Hursley, IBM and the computer industry. For IBM the existing volume display was the 3270 which contained a Motorola 16 bit microprocessor and 64 K bytes of storage. The host mainframe provided a 3270 data stream connection to the terminal that was decoded by the microprocessor. Large capacity terminal memories were not available. The memory 64 K limitation of the 3270 was a problem being, apparently, too small for user applications.

Martin Pinnell from Hursley, and Dick Talmadge from IBM Research in the USA, had long been interested in a single address space covering a terminal and the host that would handle unlimited user data and code. Martin had been working on a small operating system that would transfer data to the terminal using a Most Frequently Used algorithm. For this to work the terminal had to discard unwanted data and return needed data to the host when local memory space ran out. This idea gave the terminal unlimited memory space. The project was named the Virtual Memory Terminal (VMT).

This idea rapidly gained traction under the determined guidance of David (DAT) Reid who migrated from the discontinued State Farm activity. Brian Marks, a senior programmer started to provide the host support while Bob Maddock was enlisted to write a text processor to work in this environment. The speed with which this work was done was impressive and the text editor, which worked well, was the first demonstrable validation of the ideas. A prototype windowing system, which I developed, demonstrated the opportunities for more terminal function. The applications were written in ‘C’ which had replaced Pascal as the language of choice. In the USA architecture attempts were being made by S/360 architects to integrate the 3270 architecture into S/360 architecture.

With that in mind two high-end display development projects (Aragon and Genoa) were started in an attempt to produce products based on extending the capability of the existing 3270 terminal architecture for graphics and other uses. There was considerable rivalry between the VMT and the Aragon/Genoa developers and the 3270 architecture extension was the IBM strategy. This made it very heavy going for VMT. Subsequently the integration of the 3270 and S/360 architectures was not possible and it never happened.

Microsoft announced Windows and attempts by IBM to set the course of history with extended 3270 architecture failed. The new 3270 architecture was dropped and an IBM Windows rival in the shape of OS/2 emerged. Eventually IBM sold its PC arm. The emergence of a dominant Microsoft, lack of meaningful senior management for VMT support and the need for OS/2 manpower was fatal. Martin Pinnell implemented his plan to emigrate to Australia. The natural progression of the prototype VMT, which used existing 3270 host/terminal interchange procedures, could have met IBM needs to expand terminal capability with existing architecture and fully exploit host storage whilst being compatible with enlarged terminal memory.

These basic characteristics are similar to what is now thought of as Cloud Computing. In retrospect we now know that technology development changed all the issues.

For many years Hursley had a strong graphics capability with the host program GDDM (Graphics Display Manager) using the 3270 display. At this time Hursley was liaising with Microsoft on their development of graphics for Windows with the objectives of providing the GDDM to their operating system. There were many reciprocal visits between Seattle and Hursley which did not materialise an outcome for Hursley. The main beneficiaries from this were Microsoft who was alerted to the overwhelming need for terminal graphics capability.

Like many others in Hursley I was involved in visits to Seattle. The Microsoft Lab there had the feeling of competence and enthusiasm that was very infectious, similar to the early days at Hursley. After a technically challenging meeting it was common for them to organise a softball kick about in the corridor, to relieve tension, followed by delicious sandwiches from the local Deli! These visits to Microsoft recalled for me the exuberant memories of 6300 with exaggeration!

Bill Gates, relatively unknown to the world at that time, made visits to Hursley where he was treated with great respect. His Programmers were very creative and deserved their success.

Hursley liaison with Microsoft ceased and as a result of VMT and GDDM Hursley became a partner of the IBM Boca Raton and Austin Laboratory in OS/2 development. The emergence of intelligent terminals using Windows and other terminal operating systems had killed the 3270 architecture stone dead but OS2 was not to become one of the world’s universal operating systems.

Software was becoming increasingly important and my time was up. I had seen the clunk of relays to the internet dominance with almost perfect timing, but whatever happened to the small company idea?