2 March 2012
During the early 1950s, J. Lyons and Co., a major British corporation celebrated for its fine foodstuffs and chain of tearoom cafes, pioneered the introduction of some of the very first office application software to accomplish core business tasks like invoicing, payroll, order management, stock control and distribution, tasks which had formerly been exclusively clerical in nature. To accomplish its goal, Lyons went into partnership in 1948 with the Computer Science Laboratory at Cambridge University to build an electronic digital computer called LEO (i.e. Lyons Electrical Office). This, the LEO 1, was devised from the get-go as a copy of EDSAC, a machine nearing fruition at the Laboratory under its remarkable director, Maurice Wilkes, and soon to become the world's first fully-fledged example of the new generation of stored-program computers. But whereas EDSAC was intended primarily for scientific work, in particular the execution of successive differential equations requiring relatively little memory to process, priority was given at the outset to equip LEO 1 with a significantly expanded memory to enable the retention of the vastly greater quantity of input and output data typical of office applications. The very repetitive processing also characteristic of such applications necessitated a focus on faster and more reliable peripheral devices like tape drives, printers etc. and the development of fault-finding application software to monitor their operation. A specialist, largely autonomous, computer division under the inspired direction of John Pinkerton, a former radar colleague of Wilkes, was established within the company to manage progress on LEO. This subsequently branched out into the building, marketing and installation of LEO 1 replicas at client sites, on-the-job training in hardware operation and Lyons' own office application software being generously provided to customers by the company's technical staff to facilitate the transition to the new work environment. A new, upgraded version of the LEO 1, the LEO 2, appeared in the mid-1950s, to be followed by the LEO 3 at the end of the decade. The LEO 3, a radical departure from the LEO 2, embodied the very latest transistor technology and enjoyed considerable marketing success, the final 326 model becoming in particular a byword for reliability and state of the art functionality. Sadly, Lyons' senior management, who throughout had shown insufficient appreciation of the unique contribution of their computer operation, decided at this point to sell the division and get out of the data processing business. Some fifteen years later, Lyons itself was to disappear too as a corporate entity, a victim of systemic management inertia and the spate of business amalgamations and buy-outs of the late 1970s.
Georgina Ferry's `A Computer called Leo' is an interesting and informative account of this extraordinary venture, one surely unprecedented in the entire history of IT. However, by often failing to fully clarify the distinction between not only hardware and software on the one hand, but also between internal system software and application software on the other, she unintentionally overstates the overall significance of this achievement. Whereas internal system software is an integral, permanent and indispensable feature of the hardware environment, application software is externally developed and therefore incidental, disposable and throwaway - like clothing on the human body. As a virtual clone of EDSAC, the LEO 1 was not innovative or revolutionary from a hardware and internal system software perspective. Rather, its uniqueness lay in a) its in-house development by a purely commercial organization (i.e. J. Lyons) with no prior computer expertise and b) its speedy adaptation (again in-house by Lyons) to run a range of completely novel application software (likewise developed in-house by Lyons) to replace some of the most typical and fundamental, hitherto purely clerical, internal business operations. Instead of being the `world's first office computer' as Ms. Ferry's narrative appears to suggest, the LEO 1 is better viewed as the world's first computer to run the first office application software of this particular type. For it was the UNIVAC 1, designed from the very outset for the commercial market by Americans J. Presper Eckert and John Mauchly (whose brainchild office computing was), which truly qualifies as the world's first business computer. Fully operational by March 1951 - several months before the LEO 1 - and equipped with the very first examples of office application software to perform sorting and collating procedures, the UNIVAC 1 entered service with the US Census Bureau later that year, at almost exactly the same time as Lyons got to run its first job - Bakery Valuations - on the LEO 1. To her credit, Ms. Ferry provides a fairly detailed account of the UNIVAC 1's development but fails to properly acknowledge its trail-blazing role and significance.
A few more points. Ms. Ferry inexplicably commends famous mathematician John von Neumann for his authorship and circulation in 1945 of the `First Draft of the Report on the EDVAC', a shockingly thoughtless act which not only failed to acknowledge the primary responsibility of Eckert and Mauchly (his closest colleagues on the EDVAC project) for much of the material content of the document but was done without their prior knowledge or participation. Equally irrationally, she dismisses as `much ink and hot air' any objections to them being denied credit for this content. On the subject of flow-charting, she records the Lyons team as being the first to pioneer this design technique and states that it is still widely in use today, or at least when the book was published, i.e. in 2003. Yet flow-charting almost certainly originated in the mid-1940s work of von Neumann and his younger colleague, Herman Goldstine, and became largely discontinued as a popular system design tool by the early 1980s in the wake of third generation computer languages and newly introduced methodologies like advanced data flow diagramming, UML, entity process modelling and other CASE-based tools.
Ms. Ferry ends by speculating on the reasons for the repeated failure in Britain, especially at the governmental level, of very large, expensive computer projects and vacuously concludes this is primarily down to the failure of modern system designers - in contrast to the record of the Lyons pioneers - to remember `what the computer was for' and to make sure `it worked for business'. Let me just say this. Think user indiscipline - too many competing stakeholders, chaotic project management, little or no attempt at phased, modular approaches to implementation and, above all, uncontrolled specification creep. Case closed, end of story.