According to the great repository of technical and historical knowledge, Wikipedia, the first network e-mail was sent in late 1971. ARPANET, the network over which it was sent, was only two years old, but electronic mail soon became one of the system’s most popular services. The technical details took some time to sort out, but eventually the formats created to send ARPANET mail became what we all use today.
The ARPA in ARPANET was the Department of Defense’s Advance Research Projects Agency (in 1972 it picked up a D for Defense to become DARPA, the name it still has today), so it’s no surprise that the armed forces were involved in experimentation on the system. That included seeing how the military could take advantage of the electronic messaging format.
The Department of Defense had been operating an automated data networks for a long time: the Automatic Digital Network (AUTODIN) had been operating since 1963. But AUTODIN was essentially an organization-to-organization service. It connected offices, not individuals, who still received printed messages the old-fashioned way. Giving individual users messaging services by computer, which DARPA started investigating in 1975, would have been a new and different capability.
The result was the Military Message Experiment, which occurred in 1979. Using a variation of the e-mail protocols used on ARPANET, the experiment gave a single component on the staff of the Commander in Chief Pacific (CINCPAC) in Hawaii, the Operations Directorate (J-3), a cluster of networked computer terminals. Instead of having AUTODIN messages printed out at a central location and then sorted, filed, and circulated by the administrative office (J301), the messages were stored on a mainframe (a DEC PDP-10) and accessed from one of 25 terminals and 7 printers shared by the 200 people in the Operations Directorate.
The results were impressive. Pushing paper, J301 took roughly seven hours to process all the message traffic that arrived overnight. Using the computer, it took less than an hour and a half. The average time it took a priority message to go from the AUTODIN switch to the recipient dropped from two hours to thirty-seven minutes. Unsurprisingly, when users were surveyed they massively preferred the automated system for most purposes.
Needless to say, the biggest problem the MME demonstrated was there just weren’t enough computers for everyone. The experiment’s final report concluded that the staff of the Operations Directorate, about 200 people, needed about 60 terminals, not 25; the entirety of CINCPAC would require about 200.
Could CINCPAC have given its entire staff e-mail in 1979? The service would only have really counted as e-mail within CINCPAC headquarters, since communications outside would still have had to go via AUTODIN. And the messages still used AUTODIN’s headers, so they were technically not going to individual addresses, just being sorted and sent to the relevant action officer by the administrative staff. Still, the basic functionality was certainly there.
Unfortunately, the otherwise comprehensive final report doesn’t seem to have an estimate of the cost involved. The software itself, SIGMA, was mostly developed. Two of the programmers, in an article they published estimated that SIGMA was “about a 30 man-year program. This is by no means a gigantic effort by industrial standards, but large when judged by research community criteria.” More mainframes wouldn’t have been cheap, since the KL version of the PDP-10 used in the experiment cost $600,000, but the terminals weren’t that expensive. A contemporary ad in Computerworld says a Hewlett-Packard 2649A (the terminal of choice for the experiment) cost $2,150.
So let’s make a wild guess: five mainframes (the original number the team from USC proposed to network all of CINCPAC) and 200 terminals would cost $3.43 million. If hardware costs were only a fifth of the total cost that would mean networking CINCPAC would have cost $17 million – not that far from the $22 million the USC team proposed in 1973 for a far larger system with 2000 terminals. Adjusting for inflation, that’s about $56 million in 2015. That’s not a small cost, but it pales in comparison to actual major projects of the era.
To give one example, in 1979 the Navy was planning to spend $633.5 million to procure (not develop, procure) 439 Tomahawk cruise missiles. At $1.44 million a pop, that means the CINCPAC e-mail system would have been equal to the cost of just 12 missiles – or about a tenth of the number fired on the first day of the 1991 Gulf War. From that perspective, a CINCPAC e-mail system might or might not be worth it, but it certainly would have been relatively cheap.
A Note on Sources: The six volume final report on the Military Message Experiment is available online via DTIC. In addition, programmers involved in the project published several articles on the experiment in the proceedings of the 1979 National Computer Conference. They included “SIGMA – An interactive message service for the Military Message Experiment,” “The terminal for the Military Message Experiment,” and “The SIGMA experience – A study in the evolutionary design of a large software system.” The estimate for the projected cost of a Tomahawk missile comes from the data appendix to the RAND report The Joint Cruise Missiles Project: An Acquisition History (N-1989-JCMPO), available online from RAND here.