Pentagon to Pac-Man

This book is long overdue. Scholarship in the history of computing has tended to focus either on the development of a specific computer (for example, the Leo) or on the story of a manufacturer (such as IBM).

Historians have acknowledged the importance of software almost from the start of modern electronic computing, but they have been reluctant or unable to find ways to examine its history in a coherent way, other than through specialised histories of programming languages At the same time, paradoxically, this book could not have been written any earlier. The reason is that in the public's eye "software" is a term synonymous with Bill Gates - or at least Microsoft. Thus, although one could write a perfectly good history of software before the personal computer era, most readers would find it incomplete. The popular press is only partly to blame for the impression that Microsoft and its Silicon Valley rivals constitute the whole of modern software - the scholarly literature has not done enough to correct that. One of the strengths of this book is that Martin Campbell-Kelly manages to tell the story of both eras - the mainframe, punched-card era dominated by IBM and the personal computer era dominated by Microsoft - as a continuum, not as two separate worlds, which is as it should be.

Campbell-Kelly further divides the era of software before the PC into two epochs. The first is a period when specialised contractors wrote custom software, mainly for a US military customer that had specific cold-war needs. These contractors, such as the System Development Corporation, in essence created the software industry using US Department of Defense monies, by training programmers who later would fan out into the whole of the computing industry. These companies not only got the large federal and military systems to work, in doing so they became the "software engineering" school for the nation.

Campbell-Kelly, a professor at the University of Warwick, recognises how military support gave the Americans an advantage over the British that they have kept ever since, even if the British had an early lead in programming with the Cambridge Edsac computer and the Lyons Leo, arguably the world's first commercial data-processing system. The author does not say much about British or continental European developments because those countries simply did not play much of a role. Neither does he say much about Japanese software, other than a brief discussion of computer games in the final chapter. That, too, makes sense, although I wonder if he was aware of how much the Americans feared a Japanese takeover of the US software industry in the 1980s.

The second era that the author discusses is one of commercial packages sold to corporations or government agencies, rather than to individuals, that were meant to run on large systems, whether mainframes in the early era or workstations later on. Some of the big players in this era are still among us: Oracle, SAP (a German company and an exception to the rule of US dominance) and Computer Associates. IBM, too, makes and sells software at this level, and taken together this segment of the industry is far larger than Microsoft and its PC brethren.

The third segment is, of course, the PC era, with Microsoft the main player. But in a chapter with the telling title "Not only Microsoft", the author shows just how diverse this segment is, and how much real competition can be found among suppliers to that market. A final chapter deals with home and recreational software, which is too often neglected by analysts, even though its sales make it an enormous segment of the industry. Perhaps the neglect stems from the image that only teenage boys are interested in it, which is not entirely true.

Campbell-Kelly buttresses his analysis of each segment with a generous set of tables showing the revenues and sales figures of key players over time.

What the tables show time and again is the exponential growth of software as it became an integral part of any post-industrial economy. This growth is the counterpart of the famous "Moore's law", which states that computer memory chips double in density about every 18 months. That "law" - actually an empirical observation - has been studied ad nauseam . What is fascinating here is how the growth of software - the number and kinds of applications, the number and kinds of companies entering the field, the complexity of the programs themselves and so on - mirrors almost exactly the growth in chip density since about 1960. In hindsight, this seems obvious but I have not seen it so well demonstrated before.

My only criticism of this book is that the tables tend to overload the reader with data and probably could have been condensed or better presented in graphical form, as Moore's law is typically presented. That quibble aside, I strongly recommend this book to anyone who desires a complete picture of the explosive growth of computing in the past 50 years.

Paul E. Ceruzzi is curator of aerospace electronics and computing, National Air and Space Museum, Smithsonian Institution, Washington, DC, US.