1. Introduction

I am pleased to have the opportunity to publish this paper as a tribute to the pioneering contribution of KD Tocher in the development of simulation.

Keith Douglas Tocher (or 'Toch', as he was almost universally known) was born in 1921, gained a first-class BSc in Mathematics in 1941, an external first-class BSc in Statistics in 1946, a PhD (part-time, subject: The Design and Statistical Analysis of Experiments) in 1952, and was awarded a DSc by London University in 1957. He worked for the Ministry of Aircraft Production from 1941 to 1945 and then the National Physical Laboratories until joining Imperial College in 1949. He moved from Imperial College to the United Steel Companies Limited in 1957, and was appointed Professor of operational research (OR) at the University of Southampton in 1980. Toch was awarded the Silver Medal of the OR Society in 1967, Honorary Fellowship of the British Computer Society in 1971, and was elected President of the OR Society for 1972–1973 (Figure 1).

Figure 1.

Professor KD Tocher (1921–1981).

Full figure and legend (62K)

To recognize the originality of his contribution, we must first appreciate the historical context of the time regarding OR, computer hardware and software, and simulation itself, as well as consider the industrial stimulus to his innovations. A fuller discussion of the history of simulation can be found in Nance (1996), Nance and Sargent (2002), Robinson (2005) and Hollocks (2006a).

2. Operational research

Much has been now written about the history of OR (eg Kirby, 2003). It emerged in the 1930s through studies of the newly developed radar systems, and grew significantly with its work in support of the Army, Navy and Air Force during World War II (WWII), such as in bombing accuracy, U-boat searches, aircraft maintenance and tank gunnery. It was characterized by a multi-discipline nature, although with a strong statistical content, and the high calibre of staff. Following the end of the war, OR was taken up by major industries such as coal and steel. Specifically, we are interested in its development in one particular section of the UK steel industry, namely the United Steel Companies Limited.

In 1949, Stafford Beer (1926–2002, http://www.guardian.co.uk/obituaries/story/
0,3604,785671,00.html, accessed 28 May 2008) was appointed Production Controller for the Samuel Fox subsidiary of United Steels. Building from this, he pioneered the introduction of OR (as well as his interest in another emerging field, Cybernetics) and by 1956 had persuaded the Board of United Steels to establish a central Department of OR and Cybernetics. The following year, a building was acquired at 1 Tapton House Road, Broomhill (S10 5BY) in Sheffield as the base for this new department and it was renamed Cybor House (Hollocks, 2006b). Beer recruited strongly, seeking a high-calibre multi-disciplinary team and this opportunity proved very attractive to Tocher. In 1958, the adjacent property in Tapton House Road, known as Redlands, was added, substantially to provide accommodation for the newly acquired Ferranti Pegasus computer.

The department at this time incorporated both OR and Cybernetics, the latter being particularly driven by Stafford Beer himself. However, this emphasis came to an end from 1961 when Stafford Beer left the company (jointly founding a consultancy christened SIGMA) and was succeeded by David Owen. An increasing emphasis on OR followed and by 1964 the department was led by Toch and staffing had reached 80–90.

In addition, the Operational Research Club was founded in 1948, produced a quarterly journal from 1950, and became the Operational Research Society in 1953. By 1964, membership had reached 1242 (Cummings, 2008). OR did not feature in UK universities until programmes started at Imperial College and Birmingham University in the early 1960s. The first university OR Department, under Professor BHP (Pat) Rivett, opened in 1964 at the new University of Lancaster (where Toch gave the very first lecture).

3. Computer hardware

Calculation and computation has a long history of supporting devices. From the abacus to Jacquard's loom, Napier's bones, Pascal's calculator, Babbage's difference and analytical engines, and Hollerith's calculator and tabulator. Several books address the history of computing, historically or contemporaneously, such as Halley (2005) and Bowden (1953), Tocher contributing a chapter in the latter. The description that follows draws on them as well as more direct information.

The first material steps toward 'computers' came during WWII, but this followed design thinking in 1937–1942 by John Atanasoff and Cliff Berry at Iowa State College (now University) and around 1936 by mathematician Alan Turing, then at King's College, Cambridge. The war added urgency but little was actually completed until after 1945.

In the USA, Howard Aiken developed, with IBM funding, the Harvard Mk.1 calculator (which was electromechanical, 51 ft long and weighed 5 tons) in 1943, but the world's first programmable digital electronic computing device was the UK's Colossus. This was designed in 1943 by Tommy Flowers, an engineer at the General Post Office laboratories in Dollis Hill, London and operational in 1944. This was a valve-based electronic device but not a general purpose computer, being specifically focussed on cryptography. Ten were made by the end of WWII.

The first general purpose computer was unveiled by John Mauchley and Presper Eckert at the University of Pennsylvania in 1946 (design work having started in 1943). This was the one-off ENIAC (Electronic Numerical Integrator And Computer), a base-10 machine with 1800 valves and 70 000 resistors (Figure 2). The mathematician John von Neumann was an adviser on the project. ENIAC's first use was in part of the hydrogen bomb development project and the computer operated up to 1955.

Figure 2.

ENIAC.

Full figure and legend (96K)

Meanwhile, back in the UK, the leading work was at Manchester University who completed their Mk.1 machine in 1948. Work was also under way at Cambridge University (with a grant from J Lyons and Company Limited) on EDSAC (Electronic Delay Storage Automatic Calculator). The world's first computer in routine commercial service was then J Lyons and Company's LEO 1 (Lyons' Electronic Office), designed in 1949 and commissioned in 1951. Ferranti launched their Mk.1 general purpose computer, derived from the Manchester Mk.1 machine, also in 1951 (1951, incidentally, was the year of the Festival of Britain). A fuller discussion of the early history of computing in the UK can be found in Lavington (1980).

Following their ENIAC work, Mauchley and Eckert with von Neuman, proposed in 1949 a successor, EDVAC (Electronic Discrete Variable Automatic Computer) building on some distinctive principles: binary operation and distinction of input, processor, control, output and memory. However, due to disputes with the University of Pennsylvania over intellectual property, Mauchley and Eckert left to form the Electronic Control Company and designed the UNIVersal Automatic Computer (UNIVAC). Unfortunately, funding problems beyond their control meant that their company was taken over by Remington Rand who, again in 1951, delivered the initial UNIVAC Mk.1—the first commercial computer. In total, 25 of that model were sold. A fuller discussion of the early history of computing in the USA can be found in Rosen (1969).

In the midst of this early computer development, Toch developed his own interest in the field and joined Imperial College shortly after Sydney Michaelson (1925–1991), a numerical analyst and later Professor of Computing at the University of Edinburgh. They took over from Tony Brooker (who moved to Cambridge University and then to Manchester University) who had been working on the Imperial College Computing Engine, ICCE1, since 1947 (Lehman, 1999; Computer Conservation Society presentation available at http://www.macs.hw.ac.uk/~greg/icce/
lehman.ppt, accessed 8 March 2008). ICCE1 became operational in 1953 and ran until 1957. (Brooker later became Professor of Computing Science at the University of Essex, 1967–1988.)

Around 1952, before ICCE1 was launched, Toch and Michaelson started developing their ideas for ICCE2, with more electronic operation. Reportedly, they scoured the Tottenham Court Road and similar locations to find ex-MoD and other low-cost components, adopting an original modular approach to the architecture. Brooker observed that Tocher was critical of other designs, quoting Toch as saying 'Of course ... never aimed to build the world's best computer, but rather the world's first computer' and that Toch 'deliberately sought the perfect architecture for the limited resources at his disposal' (RA Brooker in an Introduction to Tocher's ICCE1 Report (Tocher, 1952a) available from ftp://ftp.macs.hw.ac.uk/pub/funcprog/icce/Tocher.doc, accessed 8 March 2008). However, in 1957 while the ICCE1 was operational and ICCE2 (Figure 3) under development, Imperial College stopped the funding entirely, in favour of acquiring a mainstream commercial computer—a Ferranti Mercury. This contributed to Toch's departure, accepting Stafford Beer's invitation to join Cybor House, attracted by the OR opportunity there. Allegedly Toch took some of ICCE2 with him when he moved from Imperial College to United Steels, but its subsequent fate is unknown.

Figure 3.

ICCE2.

Full figure and legend (97K)

The Mercury was only one of the Ferranti models, thermionic-valve based as other first-generation machines, and of particular note here is the Ferranti Pegasus, launched in 1956. By this time, computers had become much more commercial in appearance, construction (eg designs featuring easily replaceable boards) and production. Ferranti sold 26 Pegasus Mk.1 and 12 of the Mk.2 (launched in 1959). By this time, transistors (patented before WWII) were appearing in second-generation computer designs, for example, the Elliott Automation 803 in 1960 and 503 (60 times faster than the 803) in 1962. In 1964, the computing world changed dramatically with the announcement by IBM of the System/360 which took a more holistic view of hardware, software and applications and set the pace for third-generation computing.

One of the Ferranti Pegasus was sold to United Steels at Cybor House (Figure 4), although shared with Sheffield University, and another to the British Iron & Steel Research Association (BISRA) in Battersea. Some sources identify the United Steels' machine as a Pegasus II, but that would be inconsistent with the timing which would suggest a Pegasus I. However, hardware work carried out by Cybor House technical staff meant that it soon became non-standard anyway! Subsequent to original installation, magnetic tape drives were added (not shown in Figure 4 but located in front of the fireplace to the right). Pegasus used five-hole paper tape as its input/output medium, such tapes being prepared or printed on Creed teleprinter devices.

Figure 4.

Cybor House Ferranti Pegasus.

Full figure and legend (85K)

4. Software

A popular view is that The Countess of Lovelace, Ada Augusta Byron (daughter of the poet), was the first computer programmer, based on her 1843 written Notes on Charles Babbage' work. However, it appears (discussion chaired by Melvyn Bragg in BBC 4's In Our Time, broadcast 6 March 2008; podcast available from http://www.bbc.co.uk/radio4/history/
inourtime/rams/inourtime_20080306.ram, accessed 28 May 2008) that her main contribution was to press Babbage to actually complete his Analytical Engine!

Early programming (ie the selection and sequencing of computer operations), for example on ENIAC, involved wiring the programs across a very large plug-board (on the left-hand side in Figure 2). In due course, computers actually stored the programs (although there is some dispute over which design did that first). Computer architecture progressed using a binary basis, leaving behind ENIAC's base-10 operation. However, programming in binary code was hardly practical.

Programming then used a primitive and cryptic numeric or alphanumeric symbolic code, only one step removed from the computer's binary operation. By 1950, marginally more useable autocodes emerged starting with Univac Short Code and followed by a Manchester Mk.1 Autocode compiler in 1952. Extracts from the Ferranti Pegasus machine code and autocode are shown in Figure 5.

Figure 5.

Extracts from Ferranti Pegasus machine code and autocode, c. 1964.

Full figure and legend (23K)

This was the software environment, dominated by machine codes and autocodes, as Tocher joined the United Steel Companies in 1957. It was not until that year that, in the USA, IBM launched Fortran, FORmula TRANslator (IBM, 1957), with Fortran II in 1958 and version IV in 1962. These launched in Europe rather later. In 1960, a working group released the design of a powerful scientific language Algol60—ALGOrithmic Language (Backus and Naur, 1960). Slightly later, an international conference agreed on a specification for a Commercial and Business Oriented Language, Cobol (US Department of Defense, 1960). Subsequently, BASIC was developed by John G Kemeny and Thomas E Kurtz at Dartmouth College, Hanover, New Hampshire, in 1963 (manual available at http://www.bitsavers.org/pdf/dartmouth/BASIC_Oct64.pdf, accessed 29 May 2008) and IBM announced PL/1 (Sibley, 1965) in 1964, developed at IBM Hursley Laboratories, Hampshire, UK.

5. Simulation

Simulation has its earliest roots in the 1930s with the emergence of Monte Carlo methods, that is the use of a suitable statistical sampling process to resolve problems intractable to analytical methods. (Note: the first flight simulator patent was registered in 1929.) So, to Tocher at Imperial College simulation was already a well-established technique, although in 1952 he observed that he did 'not feel that the use of these Monte Carlo methods should be regarded as anything but a stop-gap procedure' (Tocher, 1952b)!

The initial computerized sampling of Monte Carlo models progressively gave way to more involved bespoke models of real systems but, necessarily, written in machine code or autocode level software. An example is a model of the Port Talbot Open Hearth steel plant of the Steel Company of Wales which, when reported in a conference paper in 1958, was still unfinished after 2 years (Neate and Dacey, 1958). This work used the BISRA Pegasus on a contract basis. In the USA, Julian Reitman, trying to model on an IBM 650 queuing in airline reservation systems, concluded that 'simulation is not a useful tool in the 1950s' (Reitman, 1988).

Across business and industry, manual/'hand' simulation was a not uncommon tool in OR and Work Study departments' activity in the 1950s/1960s, using tables of random numbers as the foundation. This was still the case into the 1970s (Szabo and Lyons, 1971).

The era of second- and third-generation computing was a time of bundled software, that is application software made freely available with hardware when acquired (although sub-routine libraries had been provided with first-generation machines). This included simulation software, such as the Elliott Automation Simulation Sub-program Library (Williams, 1962; Elliott Scientific Computer Division, 1964).

6. The new approach

Soon afterwards, Tocher joined the United Steels' OR Department, he was asked by Stafford Beer, to model one of their steel plants. This was not a pioneering challenge in itself as steel plant simulation had already been carried out (eg the Neate and Dacey project referred to above, which started in 1956). However, Tocher knew that United Steels had a number of steel plants across the north of England. The plants in Scunthorpe, Rotherham, Sheffield and Workington covered three different technologies: open-hearth, electric arc and Bessemer converter. (Another major technology, basic oxygen steel making, was on the horizon.) So, Tocher saw the challenge as in producing a comprehensive model that could be used for any of these sites—a General Steelplant Program, GSP. Tocher had already been researching and writing on computers in sampling experiments while at Imperial College (eg Tocher, 1954).

Although there was clearly a similarity in purpose across the United Steels' steel plants, the various technologies, equipment and layouts meant important differences in modelling. Tocher thus had to conceive a framework that would address the steel plant problem more generically.

The core idea in this came to him (evidently while in his bath!) at Christmas 1957. The notion started from the concept of a system consisting of individual components, which he termed 'machines', progressing as time unfolds through 'states' that change only at discrete 'events'. This was then subject to a three-phase process (Figure 6):

A:

Advancing time to that of the next scheduled event(s), that is, the one or more 'Bound' to occur at that time;

B:

processing the B-event or events arising; and

C:

processing other prospective events whose occurrence is Conditional on certain defined circumstances—which may have been influenced by the just-completed B-phase.

Figure 6.

The three-phase structure.

Full figure and legend (34K)

Then returning to the A-phase.

Toch argued that 'the B and C phases of GSP represent the event and activity phases respectively' (Tocher, 1966b). A real-world system could then be fully described by event/activity descriptions processed under the control of an 'Executive' managing the three-phase process.

As this idea was developed into computer software, it became clear that GSP was not simply a General Steelplant Program but was a more powerful framework and could be regarded as a General Simulation Program.

Work on GSP, written in machine code, became a key task on the new Cybor House Pegasus computer. Some elements of GSP design were influenced by the Pegasus hardware architecture, in particular the central data structure, the 'Matrix', an 8 n array the columns/vectors of which were designated 'S' to 'Z'. The leading rows of the array (1-m) were assigned to represent the time-dependent machines of the model being built, the value of m being specified by the model-builder who could identify each row with such machine of the problem as they thought appropriate. The T-vector across these rows/machines was used by the Executive to hold the times of the next events for each machine. By convention, the 'S' vector of the time-dependent machines was used by the model-builder to hold the current state of each machine. The remaining columns/vectors of the first m rows were available for machine-oriented data, for example, parameters for sampling distributions. The subsequent rows of the Matrix could be used for non-time-dependent entities, that is, stock holdings, or general data storage for the model. This proved a very useable structure, but the choice of an eight-column matrix was a function of the Pegasus architecture, which managed data in computing memory in eight word blocks (Felton, 1962). This aspect of the design outlived the Pegasus implementations, on to the 503 and to FORSS.

7. Early developments

Only one simulation model was built using the pilot version (Mk.0) by Peter Amiry in 1958 (Hollocks, 2006b). The first substantive version of GSP, Mk.1, was released in 1959 (Tocher et al, 1959) with a subsequent volume 2 to the manual (Tocher, 1961). The language pioneered not only discrete-event simulation but also the concepts of block structure and even objects. Given the hardware and software context of the time, GSP was an extraordinary innovation. It was made public in the seminal paper by Toch and David Owen presented at the Second IFORS Conference (Tocher and Owen, 1960).

The spur for Toch's innovation was a steel plant problem with the intrinsic cyclical nature of the central steel-making process. It is interesting to reflect whether Toch's basic concept would have been any different had his first problem for simulation been a rolling mill or transport problem—or even something outside of the iron and steel industry, say in automobile manufacture or airports.

Although the three-phase process was Toch's pivotal innovation for simulation software, he argued that 'the first step in writing a simulation is to prepare a flow diagram' (Tocher, 1960). His approach to this was itself distinctive taking a form which highlighted the activities of the system studied and hence potential 'B' and 'C' events. His illustration from that 1960 paper (which, incidentally, predates the Tocher and Owen IFORS paper by several months and itself refers to GSP) is shown in Figure 7. Toch included the approach in the GSP Mk.2 manual (Tocher and Hopkins, 1964) where it is referred to as a 'wheel diagram', the term commonly used at Cybor House, and discussed the approach further in a paper at an IBM symposium in 1966 (Tocher, 1966a). It is now usually referred to as an Activity Cycle Diagram and the concept has had its own continuing life, for example, Hlupic and Paul (1994), Martinez (2001) and Araujo and Hirata (2004).

Figure 7.

Illustration of Wheel/Activity Cycle Diagram (from Tocher, 1960).

Full figure and legend (73K)

Another associated innovation at this time was real-time data collection for simulation, referred to in both Tocher and Owen (1960) and Tocher (1960). The work at Cybor House in this area used semi-automatic plant performance recording equipment including the 'UniSteel Automatic Recorder'. It was deployed in some projects but use lapsed. At this time, Tocher also anticipated 'experimental goal searching' in simulation using 'the Box evolutionary technique which makes a systematic search of the (parameter) phase space' (ibid.). It is not known how far that plan progressed.

It is of note that iron and steel industry processes include some that are continuous, for example, the blast furnace, and others that are part continuous, for example, the electric arc furnace. The accommodation of continuous elements as well as discrete within simulation models was never regarded as a problem by Toch. For example, a Bessemer process control project started in 1961 involved a mixed discrete-continuous test-bed simulation model (Tocher and Splaine, 1966).

8. GSP evolution and other software

Subsequent to Tocher's work on GSP Mk.1, a team in the BISRA produced a package called Montecode (Kelly and Buxton, 1962) on their Ferranti Pegasus. The BISRA team included John Buxton who had worked with Toch on GSP Mk.1. Work on simulation packages appears to have started in the USA around 1960 (Nance, 1996) with Gordon and others. Markowitz of Remington Rand produced Simscript as a platform for work in warehousing and inventory management (Markowitz et al, 1962), Kiviat produced GASP as a basis for addressing a steel plant problem (Kiviat, 1963) and Gordon produced GPSS at IBM (Gordon, 1962) to address problems within computer systems design. Simscript, GASP and GPSS followed the emergence of Fortran as an end-user programming tool and all started their own development paths. GPSS was made available by IBM as bundled software with its computers and became widely used.

In 1963, Tocher published the Art of Simulation, the first book in this field. Apparently, it had to be written a second time as the original and first proofs were lost in a fire at the publishers! It was 3 years until the next book on simulation (Naylor et al, 1966).

In 1964, GSP Mk.2 was released (Tocher and Hopkins, 1964), and later that year list processing features were added. An example of GSP-2 code is shown in Figure 8. Also, later in 1964 Cybor House installed an Elliott 503 (Figure 9) to replace the Pegasus, but initially using the old Pegasus magnetic tape drives. Tocher had to design and lead the building of an operating system for the new machine since one was not forthcoming from the manufacturer. The 503 was then used as the platform for the development of GSP Mk.3 which was released in 1967 (Tocher, 1967). An interesting dimension at this point was the recognition by Tocher and the team that the GSP design could be seen as part of a wider concept which he termed PLUS (Programming Language United Steel).

Figure 8.

Example of GSP Mk.2 code.

Full figure and legend (29K)

Figure 9.

Cybor House Elliott 503.

Full figure and legend (75K)

By this time, in Norway, Simula had emerged, progressively developing its Object Oriented (OO) features, for example, the object and class/sub-class concepts emerging in 1967 (Nygaard and Dahl, 1978). The GSP team had made little of the OO implications of their work. By this point also, the availability of Fortran and Algol meant that simulation tools could be produced by OR groups within individual companies. A notable instance was in ESSO Petroleum where a team, involving John Laski and John Buxton—both of whom had worked at United Steels in the GSP developments (and Buxton on Montecode)—and engaging Toch as an advisor, produced CSL, Control and Simulation Language (Buxton and Laski, 1969). This then had its own development path.

Adding to the pioneering work in software, there was accompanying pioneering in the simulation domain at Cybor House, in particular in visual displays and interactivity (Hopkins, 1965). Figure 10 shows a purpose-built display to portray, on-line to a GSP model, electricity power consumption in an Electric Arc plant. This formed part of a study in Maximum Demand tariff operation. Other studies involved the notion of interactive production gaming (Mellor and Tocher, 1963), for example Figure 11 shows a game concerned with steel plant scheduling.

Figure 10.

Model-driven display for a maximum demand power supply study.

Full figure and legend (90K)

Figure 11.

A production game in steel plant scheduling.

Full figure and legend (70K)

1967 saw the nationalization of United Steels as part of the British Steel Corporation (BSC), along with The Steel Company of Wales, Dorman Long, English Steel and others. Subsequently, in the 1970s, there was a change in the BSC computing environment. The growth in data processing across the organization meant a corresponding growth in the profile of that part of Management Services. Accompanying that growth, an IT strategy was shaped based on hardware and systems software from two suppliers, IBM and ICL. Up to this part of the story, scientific computing capacity had been available to Tocher's OR group, that is, the Pegasus and 503 computers. By 1970, the 503 machine had become increasingly unreliable and a review of scientific computing provision carried out within British Steel concluded with a proposal to acquire a Burroughs 7700 machine. However, the cost of this was ultimately rejected in favour of a focus of finance on the IBM/ICL axis—although a budget was provided for OR to have remote access to a Univac 1100 series bureau computer at Scicon Limited of Milton Keynes, as well as to BSC ICL 1900 equipment.

In this context of a dying 503 and the rationalization to ICL/IBM mainframes, work on a Fortran-based version of GSP Mk.3 was started in 1970 using the ICL 1905F service (Hollocks, 1971). It was ported in 1972 to IBM 360 (in BSC's Teesside& Workington Group) for further development, identified later as FORSS (British Steel, 1975a). It became used across a number of British Steel OR sites and later (back at Cybor House) formed the basis of the visual interactive simulation package FORSSIGHT (Hollocks, 1983).

Around 1970, Toch moved from his Cybor House, Sheffield base to Birmingham, to become part of a BSC Head Office Management Services co-ordination team. However, work continued in the 1970s at Cybor House on a GSP Mk.4, seen then as a dialect of a Language for OR in British Steel—LORBS (Tocher, 1979a). This involved more pioneering work, in this case in the design and development of a machine-independent low-level language, MILL (Tocher, 1979b), in which LORBS/GSP-4 were in turn written. GSP Mk.4 was launched in 1976 in pilot form (Bent, 1976), based primarily on an ICL system, although potentially portable via MILL to other equipment. A full version of GSP-4 never materialized before Toch left the Corporation. In 1975, a side project was carried out to produce a mini-computer-based software package using the computer as an executive driver based on GSP principles as support to simulation exercises otherwise carried out manually (British Steel, 1975b). The evolution path of GSP is summarized in Figure 12.

Figure 12.

GSP development path.

Full figure and legend (16K)

Through the 1970s, Toch's attention had moved beyond simulation to the notion of cost-based planning models, commonly known in British Steel at the time as 'economic models'. In the manner of producing a framework for simulation, Tocher developed with his team a framework for economic modelling in BSC with a corresponding language, the Language for Economic Modelling in British Steel—LEMBS (Tocher, 1976), like GSP-4 seen as a dialect of LORBS. However, he never lost his interest in and engagement with simulation. Toch left British Steel for the University of Southampton in 1980.

The turn of the decade saw a sea change in computing available to OR generally. The microcomputer emerged, first 8-bit and then 16-bit, with software making it a very cost-effective and usable tool, readily accessible—in particular, not requiring links to mainframes. This tool was extensively taken up by OR groups, not only in British Steel but also across the UK, for a range of applications including simulation (eg Ranyard, 1981; Hollocks, 1983). Sadly, Toch died in 1981 before he could exploit this new opportunity.

9. Conclusion—innovations and reflections

The notion of Monte Carlo modelling was, as discussed earlier, established before Tocher's involvement but he was a pioneering part of the thrust to take it beyond Monte Carlo statistical sampling to more comprehensive modelling. The pivotal innovation in his work is widely recognized as the three-phase structure with its associated constructs, providing the basis of many simulation packages used today. But the corresponding framework he established also provided the first language to contribute structure (albeit not in the sense of Algol60) and introduced a concept of objects, although they were not known as that at the time. Other innovative work included Activity Cycle (or Wheel) Diagrams, mixing discrete and continuous simulation, the notion of a visual interface to simulation, interactive models, and operational gaming—as well as an operating system. Other innovations, for example, simulation optimization, were anticipated but not (apparently) completed. Even this list omits further innovations and achievements, for example, Toch was the first industry-based visiting professor in a UK university.

There are also other principles that Toch employed as an individual. For example, he looked for the generalization from specific issues to a more generic picture, such that any solution would have wider applicability. He sought the fundamental structure of problem as a basis for its solution, creating tools that added leverage for users—exploiting technology. Tocher never saw an incongruity in a highly qualified Head of Department being involved in the real detail of a solution, even at coding level. He recognized that what is possible at that level could lead to more efficient and effective products.

At a personal level, Tocher pursued the best with integrity. Despite his own considerable intelligence and qualifications, his modesty led him to regard intelligence as material only if it was sensibly applied, and he would accept such input from anyone.

Simulation development continued beyond Toch's early death in 1981, but his legacy lives on.

References

1. Araujo WLF and Hirata CM (2004). Translating activity cycle diagrams to Java simulation programs. 37th Annual Simulation Symposium of IEEE Computer Society, pp 157–166.
2. Backus J and Naur P (1960). Report on the algorithmic language ALGOL 60. Numer Math 2: 106–136. | Article |
3. Bent ME (1976). Pilot GSP IV Manual. British Steel Corporation Management Services, Sheffield.
4. Bowden BV (ed) (1953). Faster than Thought: A Symposium on Digital Computing Machines. Pitman: London.
5. British Steel (1975a). FORSS—A simulation system based on Fortran. British Steel Corporation, Teesside and Workington Group Systems and OR Department Report, OR67/1.
6. British Steel (1975b). HISLABS computer operating instructions. British Steel Corporation, Internal Report.
7. Buxton JN and Laski JG (1969). Control and simulation language. Comput J 5: 194–199.
8. Cummings N (2008). A little OR history, 1937–1994. Inside OR January: 16–17.
9. Elliott Scientific Computer Division (1964). Elliott Simulator Package (ESP). pub TIS-503-2.
10. Felton GE (1962). The Pegasus Programming Manual. Ferranti Ltd: London.
11. Gordon G (1962). A general purpose systems simulator. IBM Syst J, 1: 18–32.
12. Halley M (2005). Electronic Brains: Stories from the Dawn of the Computer Age. Granta Books: London.
13. Hlupic V and Paul RJ (1994). Simulation modelling of flexible manufacturing systems using activity cycle diagrams. J Opl Res Soc 45: 1011–1023.
14. Hollocks BW (1971). A basic implementation of the General Simulation Program. British Steel Corporation, Special Steels Division Management Services Centre Report 211/OR 40/1.
15. Hollocks BW (1983). Simulation and the micro. J Opl Res Soc 34: 331–343. | Article |
16. Hollocks BW (2006a). 40 years of discrete-event simulation—A personal reflection. J Opl Res Soc 57: 1383–1399. | Article |
17. Hollocks BW (2006b). A glimpse of early OR—A brief history of Cybor House. OR Insight 19: 18–25.
18. Hopkins DA (1965). Control and display facilities for simulation games. In: Hollingdale SH (ed). Proceedings of the NATO Conference Digital Simulation in Operational Research. English Universities Press: Hamburg, pp 347–356.
19. IBM (1957). Programmer's Primer for FORTRAN Automatic Coding System for the IBM 704. IBM Corporation: New York, Form No. 32-0306.
20. Kelly DH and Buxton JN (1962). Montecode—An interpretive program for Monte Carlo simulations. Comput J 5: 88–93.
21. Kirby M (2003). Operational Research in War and Peace—The British Experience from the 1930s to 1970. Imperial College Press: London.
22. Kiviat PJ (1963). GASP—A General Activity Simulation Program. Applied Research Laboratory, US Steel Corporation: Monroeville, PA.
23. Lavington S (1980). Early British Computers. Digital Press: Bedford, MA.
24. Markowitz HM, Hausner B and Karr HW (1962). Simscript: The simulation programming language. Rand Corporation Report, Rm-3310, Cambridge, MA.
25. Martinez JC (2001). EZStrobe—General-purpose simulation system based on activity cycle diagrams. In: Peters BA, Smith JS, Medeiros DJ and Rohrer MW (eds). Proceedings of the 2001 Winter Simulation Conference. IEEE: Washington, DC, pp 1556–1564.
26. Mellor P and Tocher KD (1963). A steel-works production game. Opl Res Q 14: 131–135. | Article |
27. Nance RE (1996). A history of discrete-event simulation programming languages. In: Bergin TJ and Gibson RJ (eds). History of Programming Languages Vol. II ACM Press and Addison-Wesley Publishing Company: New York, pp 369–427.
28. Nance RE and Sargent R (2002). Perspectives on the evolution of simulation. Opns Res 50: 161–172. | Article |
29. Naylor TH, Balintfy JL, Burdick DS and Chu K (1966). Computer Simulation Techniques. Wiley: New York.
30. Neate R and Dacey WJ (1958). A simulation of melting shop operations by means of a computer. In: Proceedings of BISRA Conference Digital Computers in the Steel Industry, BISRA Report OR/CA/4/58. British Iron & Steel Research Association: Park Lane, London W. 1.
31. Nygaard K and Dahl O (1978). The development of the SIMULA languages. ACM SIGPLAN Notices 13: 245–272. | Article |
32. Ranyard J (1981). Introducing the micro computer into the OR department. J Opl Res Soc 32: 277–296. | Article |
33. Reitman J (1988). A concise history of the ups and downs of simulation. In: Abrams M, Haigh P and Comfort J (eds). Proceedings of the 1988 Winter Simulation Conference—San Diego. IEEE: Piscataway, NJ, pp 1–6.
34. Robinson S (2005). Discrete-event simulation: From the pioneers to the present, what next? J Opl Res Soc 56: 619–629. | Article |
35. Rosen S (1969). Electronic computers: A historical perspective. Comput Surv 1: 7–36. | Article |
36. Sibley RA (1965). A new programming language: PL/I. Proceedings of the 20th National Conference, Association for Computing Machinery, pp 543–563.
37. Szabo LJ and Lyons D (1971). Manual simulation technique. Work Study Mngt Serv, March: 196–198.
38. Tocher KD (1952a). Report on the Work of the Computer Group. The Imperial College of Science and Technology Mathematics Department: London.
39. Tocher KD (1952b). Report of a meeting on Marshalling and Queuing. Opl Res Q 3: 12–13.
40. Tocher KD (1954). The application of automatic computers to sampling experiments. J R Stat Soc Ser B (Methodol) 16: 39–75.
41. Tocher KD (1960). An integrated project for the design and appraisal of mechanical decision-making control systems. Opl Res Q 11: 50–65. | Article |
42. Tocher KD (1961). Hand book of the general simulation program. Vol. 2, United Steel Companies Ltd, Department of Operational Research and Cybernetics Report 88/ORC.3/Tech.
43. Tocher KD (1963). The Art of Simulation. English Universities Press: London.
44. Tocher KD (1966a). Some techniques of model building. Proceedings of the IBM Scientific Computing Symposium on Simulation Models and Gaming IBM, White Plains, NY.
45. Tocher KD (1966b). The structure in simulation languages; letter commenting on article by JG Laski. Opl Res Q 17: 99–100. | Article |
46. Tocher KD (1967). PLUS/GSP III Specification. United Steel Companies Ltd, Department of Operational Research, Sheffield.
47. Tocher KD (1976). LEMBS—Language for Economic Modelling in British Steel. Head Office Management Services, British Steel Corporation, Sheffield.
48. Tocher KD (1979a). LORBS—Language for Operational Research in British Steel. Head Office Management Services, British Steel Corporation, Sheffield.
49. Tocher KD (1979b). MILL Manual. Head Office Management Services, British Steel Corporation, Sheffield.
50. Tocher KD and Hopkins DA (1964). Handbook of the GSP Mark II. United Steel Companies Ltd, Department of Operational Research, June, 118/ORD/10/TECH.
51. Tocher KD and Owen DG (1960). The automatic programming of simulations. In: Banbury J and Maitland J (eds). Proceedings of the Second Conference of the International Federation of OR Societies, Aix-En-Province. English University Press: London, pp 50–60.
52. Tocher KD, Owen DG and Cunningham Green RC (1959). Hand book of the general simulation program. Vol. 1, United Steel Companies, Department of Operational Research and Cybernetics Report 54/ORC.3/Technical.
53. Tocher KD and Splaine M (1966). Computer control of the Bessemer process. J Iron Steel Inst 204: 81–86. | ChemPort |
54. US Department of Defense (1960). Cobol, Initial Specifications for a Common Business Oriented Language. #1960 0-552133, US Government Printing Office, Washington, DC.
55. Williams JWJ (1962). ESP: The Elliott simulator package. Comput J 5: 328–331.

Acknowledgements

This paper was first presented on 1 April 2008 at SW08, the Fourth OR Society Workshop on Simulation held in Redditch, Worcs, UK. Figure 2 is a US Army photo c. 1947–1955, photographer unknown; 'the image is a work of a US Army soldier or employee, taken or made during the course of the person's official duties and, as a work of the US federal government, the image is in the public domain'. Figure 3 is used courtesy of Dr David Tocher. Figures 4 and 7, 8, 9, 10 and11 are used by past permission of the British Steel Corporation. It would also be appropriate here to acknowledge the contribution of the many staff who assisted Toch in the design and development of GSP, such as Gil Bowling, Brian Nabb, Peter Amiry and Peter Mellor.