ENIAC in Action: Making and Remaking the Modern Computer by Thomas Haigh, Mark Priestley, and Crispin Rope (review)

Today, a book on one machine is something one would look at Argus-eyed. Indeed, contemporary historiography has a tendency to move away from machine-centered narratives and focus instead on the human side of computing. As a consequence, the history of computing is often less about the technical artifacts than it is about sociological, political, and industrial developments. This, however, need not imply a neglect of the actual practices around a machine. Indeed, in the past years researchers have re-engaged with the technical histories underpinning computational practices in order to integrate different approaches.

ENIAC in Actionis an important product from that perspective, and it adopts a pluralistic method. It explores the formal and engineering practices around ENIAC, which are reconstructed from a rich collection of archival sources, but also engages, for instance, with literature on gender and on military procurement. In that way, the authors carefully unravel ENIAC's history against some popular narratives and so provide a more complete image of the machine's role in history and historiography. Several persistent images or "myths" of ENIAC are deconstructed. For example, the book shows that conditional branching was not an afterthought but part of the original design. In their attack on this "myth," the authors disclose important material showing that the idea of a "programmable" machine was certainly not an afterthought. Secondly, they attack the image of the six ENIAC women as the first programmers, and in so doing argue that one does not do justice to the diversity of practices and people (male and female) around ENIAC if one presents it as the machine of a certain individual or group. Finally, they critique the image of ENIAC as the "obligatory passage point," a point of converging pasts and diverging futures, as well as the related narrative of "the" first computer.

The discussion of the EDVAC design—classically seen as the blueprint of the modern, "stored-program" computer—is nuanced and contextualized. The authors show how that design is rooted in the experiences with (original) ENIAC and the demands of a new set of so-called Monte Carlo problems to be set up on a "converted" ENIAC. Moreover, the classical notion of the "stored-program computer"—which is, as is argued, a concept that emerged only in the early 1950s—is replaced by a set of three paradigms, most notably the modern code paradigm, which can be reconstructed from the EDVAC report. In their detailed study of the Monte Carlo problems, they show how a programming "style" is developed that implements the modern code paradigm and uses the flow diagram notation of John von Neumann and Herman Goldstine. From that perspective, these Monte Carlo programs are the first modern programs. While one can criticize replacing the stored-program concept with another in its usage to identify another "first" (the first program ran in the new paradigm), the authors certainly show the need for a historically anchored reflection on some of the fundamental concepts of computer science.

In their attack on some of ENIAC's myths, the authors regretfully seem to create or confirm some others. Most significant is that the book, despite its novel insights and methods, ultimately reads like the Goldstine-Burks narrative in which most credit is given to von Neumann at the expense of others. Indeed, even though due attention is given to some "forgotten" heroes of ENIAC—in particular, Klara von Neumann—it is "Johnny" who remains the true genius at the end. This is apparent, for instance, in chapter 6 on EDVAC where the EDVAC design is described as "radical minimalism, akin to high modernist architecture" and ENIAC becomes "a gothic sprawl."

Related to this, there is a tendency toward overestimating the converted ENIAC (and so the idea of another "first") at the expense of the original. More care for the (literature on) programming practices around the non-converted ENIAC would...