Software-Engineering the Internet of Things

ENIAC WAS BUILT during the Second World War, from 1943 to 1945. Many consider it the rst electronic, general-purpose, large-scale digital computer. Picture it as a room encompassing 36 racks, three printer panels, a card reader, and a card punch. Each rack used hundreds of vacuum tubes to perform a speci c function. Many racks acted as accumulators: they received pulses corresponding to the digits of a decimal number and increased accordingly the number stored in them. Others were more specialized; a 1944 oor plan has racks labeled multiplier, partial product, square rooter, denominator, multiplicand, and so on. Three function table racks, initialized through switches, could be carted around on wheels. Today we build, connect, and congure most Internet of Things (IoT) systems by linking together their sensor, actuator, and computing nodes through cloud infrastructures, mobile apps, and the sharing of security credentials. Similarly, ENIAC was programmed by setting up function tables and switches and connecting its units in the manner required for solving a particular problem, such as generating sine and cosine tables or computing artillery trajectories and shock wave re ections. As you can imagine, such programming was time-consuming and error-prone. Then, in 1948 a remarkable thing happened. Inspired by the design of EDVAC (Electronic Discrete Variable Automatic Computer), discussed over summer school lectures at the University of Pennsylvania’s Moore School, ENIAC’s designers realized they could wire it in a revolutionary way. The wiring wouldn’t solve a particular numeric program. Instead, the designers would repurpose some of ENIAC’s accumulators so that it would read instructions prescribing what actions to perform from its numeric function tables. Think of this as building a command interpreter by assembling together already existing discrete electronic components. Thus, the new wiring transformed ENIAC into a versatile storedprogram computer. Rewiring IoT infrastructures into a generalpurpose computing fabric can similarly change how modern computation interfaces with our environment.