Harvard pioneer of digital frontiers

Earlier this year, the 50th anniversary of the Edsac computer was celebrated at the University of Cambridge Computer Laboratory. Edsac was one of the earliest "stored-program" digital computers to have an internal store or memory that could be altered and that held both the instructions (program) and data. It could also conditionally branch depending on the results of previous computation. These two important features made it very similar in operation to many of the computers we still use today. The mercury delay-tank technology developed for the Edsac helped make the internal storage possible.

Although the Edsac was among the first "real" stored-program computers in the modern sense, and may be considered the first to be put into regular operational service, there were predecessors that influenced its design. Wilkes attended the famous Moore School lectures on "The theory and techniques for the design of electronic digital computers" organised by the computer pioneers John Mauchly and J. Preper Eckert, held in 1946 at the University of Pennsylvania in Philadelphia. This was to sow the seeds of many of the ideas Wilkes used in the Edsac's design.

Two of the Moore School lectures were given by Howard Aiken of Harvard Uni-

versity. He had been considering the design of automatic calculating machines and wrote a seminal proposal in 1937 on the mathe-

matical features required in such a machine for scientific calculations. Through collaboration with IBM, he designed the Harvard Mark I computer, largely out of existing relay-based components available from

IBM for use in their existing calculators.

His insight was that by automating the sequence of calculations, the process could be speeded up very significantly compared with hand operation of machines by human

"computers".

However, the machine still read its program directly from a paper tape and could only conditionally stop. It could not branch to a particular location in the program automatically. Repetition could be achieved by literally producing a program "loop" from the tape, but it was important to ensure that this was not connected as a Mobius strip (although this could be corrected in real-time if spotted before the program completed the first iteration of the loop).

Until now there has been no significant scholarly volume on Aiken (1900-73) on a par with Andrew Hodges's widely admired Alan Turing: The Enigma , for example. Now we have two volumes that could claim to be of this calibre. I. Bernard Cohen is an emeritus professor of the history of science at Harvard University and has produced two works to match his credentials on the early history of computing. He had direct access to Aiken and held an extensive interview towards the end of Aiken's life, which adds authenticity to the books.

The first volume, Portrait of a Computer Pioneer , is written by Cohen alone; it gives a fascinating and detailed account of Aiken's life and work. Aiken had a somewhat troubled childhood. His father was an alcoholic who deserted the family home when Aiken was 12 years old. Aiken had to take on early responsibility supporting his mother and this may have reinforced his determination to succeed later in life. He subsequently studied physics at the University of Chicago and undertook graduate studies at Harvard.

During the main part of his professional career, Aiken established the Comp Lab (Computation Laboratory) at Harvard and built a succession of machines from the Harvard Mark I (operational 1944-59) to the Mark IV. The later machines had little impact on the mainstream design of computers, largely due to Aiken's conservatism in design, but this approach probably contributed to the successful construction of the earlier machines. Often using well-established components to produce state-of-the-art technology is a good strategy to ensure success. Witness, for example, Seymour Cray's approach to producing the world's fastest supercomputers in his heyday.

The second volume, Makin' Numbers , takes its title from a favourite phrase of Aiken when describing his calculating machines. This is a collection of essays by those who worked with and knew Aiken. These range from studies of his machines, educational issues and more personal recollections to some of Aiken's own writings.

Aiken's strong character pervades both volumes. His naval associations (he was a Commander after his war work in applying the Mark I to calculations for the US Navy) obviously affected the whole style of operations at the Comp Lab, which was run in a very autocratic manner. Grace Hopper, also in the US Navy and another well-known computing pioneer, worked with Aiken, thrived in the environment and went on to produce the early business programming language Cobol.

However, others at Harvard were suspicious of the academic respectability of this new applied science because the university's strengths were traditionally in the purer science and arts fields. Aiken, just as many do today, had trouble raising enough money to fund the expensive activities of his lab. Harvard gave comparatively little, but his naval and industrial contacts stood him in good stead, and his perseverance won him many substantial contracts.

Aiken could reject people even more decisively than those he embraced, and often cruelly. As well as those he considered sub-standard workwise, it is unfortunate that he fell out with the equally strong personality Thomas J. Watson, Sr of IBM, over the allocation of credit at the dedication of the Harvard Mark I in 1944. While Aiken was responsible for the design of the machine's architecture, IBM actually constructed the machine and gave a substantial $100,000 donation as well.

An aspect of Aiken that is perhaps forgotten by many is that he was a superb teacher. Makin' Numbers includes a first-hand account of this from the receiving end, which I would recommend for reading by all computer-science lecturers. By the mid-1950s he had a computer science programme in place that would be almost entirely recognisable by university teachers today.

Aiken's views were sometimes visionary, but at other times very conservative. He realised that substantial miniaturisation of computers was inevitable. However, he was also convinced that decimal operation was superior to binary to avoid conversion problems, a view that Wilkes bravely contested with him even though Aiken was very much his senior and well established at the time.

Ultimately, Aiken proved to be a one-man show. After his early retirement from Harvard in 1961 to become a business consultant, computer science at the university declined rapidly. Other schools such as the Massachusetts Institute of Technology and Stanford are now much more important forces in American academic computer science. Harvard's most significant claim to computing fame since Aiken is arguably that Bill Gates "dropped out" of the university to found Microsoft.

Aiken certainly deserves an important place in the early history of computing as a charismatic and individualistic pioneer. He was an excellent organiser and leader who achieved results in circumstances where most others would have floundered. These volumes provide a fitting testimony to his contribution to the establishment of computer science. I would recommend them to anyone with an academic interest in the foundations of modern computing history who also enjoys a good story.

Jonathan Bowen is lecturer in computer science, University of Reading.