This exceptionally interesting paper describes the four computers designed by Howard Aiken and built at the Harvard Computation Center during the 1940s and 1950s. The machines were perhaps not quite state-of-the-art, but it must be remembered that Aiken was virtually the only computer designer of that era whose machines were completed on time.
The Mark III was Aiken’s first venture into effective, large-scale storage, which he achieved with a large magnetic drum system. All of this is most interestingly described in Lee’s paper, which also contains a number of photographs that I had not seen before. Lee explains that archeological techniques were applied in an attempt to find the remains of the Mark III, which had been sold by the armed forces (for $60!), stored in a barn, and sold off as components, with the remains eventually buried in a ditch. Such was the sad fate of a historical artifact.
I have only one cavil about the paper, and it is not the author’s fault: the table “Survey of Large-Scale Digital Computers circa 1950” shows my own first machine, the ARC, as having a 20-millisecond addition time. This is incorrect. The ARC was a parallel, binary relay machine with a 21-channel parallel magnetic drum. The high-speed relays had a reliable reaction time of well under 1 millisecond, and the adder had anticipatory carry, so the addition time was 1 millisecond.