The Computer from Pascal to Von Neumann [Englisch] [Taschenbuch]

A review by Frederick A. Ware

This book contains a great deal of computer history, particularly of the period during and shorty after World War II when the critical details of stored program digital computers were finally worked out and implemented. Goldstine also has a unique perspective on these developments because of his work as a senior programmer on the ENIAC.

Unfortunately, major areas of controvesy arose once the ENIAC was completed. One had to do with the credit for the concept of placing a computer's program in the main store along with the data it was to use. This would permit a program to be easily and quickly modified - the ENIAC required cables to be moved in order to change the sequencing of its arithmetic units when a new set of ballistic tables were to be generated. This would also permit a program to process other programs, leading to the development of assemblers and compilers.

History gives sole credit for this idea to John von Neumann because most people prefer history to be simple and events to be tied to single individuals. The details of the ENIAC project support the position that Mauchly and Eckert (the ENIAC developers) should be given equal credit for the stored program concept. Other historians (and computer scientists of the period) share this view.

Goldstine professes to be neutral, but in fact is significantly biased toward von Neumann in this matter (the title of the book speaks for itself), and that detracts from what is otherwise a very readable and very entertaining book.

The book divides computer history into three eras - pre-WWII, WWII, and post-WWII. Again, this division is probably due more to the fact that the period in which the author made his most significant contributions was World War II.

The first section begins with the development of mechanical adding machines. Pascal invented one of the first such machines, hence the rest of the book's title. Other computing intruments included the planimeter, built from the two-disk integrator.

Some of the first section is also devoted to Babbage's difference engine, designed to the generation of tables from difference equations, and of his analytical engine. Boole and his development of boolean algebra (with its eventual application to digital computer design) is also covered.

Beginning in about 1900, significant computing machines were developed. The book describes Hollerith's card tabulating machines, first used in the 1890 census. It also covers the differential analyzer, an analog computer used to solve ordinary differential equations developed by Vannivar Bush at MIT in the 1930s.

The Automatic Sequence Controlled Calculator (also called the Harvard Mark I) is a digital computer with mechanical arithmetic units and a paper tape driven sequencer. Of all the early machines, this one is probably comes closest to implementing the Babbage's analytical engine.

Stibitz of Bell Labs built a series of digital computers using relays as the logic elements. The largest has 9000 relays. These machines are about six times faster than the mechanical machines.

The second section is principally a discussion of the ENIAC development. All of these early computers were really just souped-up calculaters with some kind of automatic sequencing capability). The ENIAC was the first vacuum tube digital computer. The Eniac was the most significant because of its blinding speed - the electronic components gave it a 1000x performance advantage over the other technologies. The key contribution of Mauchley and Eckert was to prove that large numbers of vacuum tubes could be operated reliably. The architecture of the machine was not significant, except to serve as an example of how not to do it in the furure.

The third section covers the post-war era. The development of large, fast main storage is the critical problem to be addressed in the late 1940s. The two principle alternatives to vacuum tube flipflops are mercury delay lines and electrostatic storage on a CRT . Both are volatile and require refreshing techniques. Both go on to be used in a number of computers in the next five years until ferrite core memory is developed.