I'm an enthusiast and writer on the history of military communications technology, particularly during the Cold War. The internet is very much part of this stor

I'm an enthusiast and writer on the history of military communications technology, particularly during the Cold War. The internet is very much part of this story, and I am asked about this controversy from time to time. The problem is, I find that people who argue for both positions are becoming much too fixated on the idea that there was some single set of influences on a complex project. There isn't really any answer to "was the internet designed to withstand nuclear war" for the same reason that there isn't really any single answer to any question about the historical motivations of complex undertakings. That's just not how history works. There are some facts which we know to be true: 1) Various components of the defense complex were actively researching survivable C2 communications, particularly beginning in the 1950s although there were earlier precedents. Many of these efforts involved ideas that were similar to those used in modern computer networking, and they sometimes culminated in built systems with meaningful similarities to the internet, like AUTODIN. 2) A diverse cast of academics, contractors, and government entities were involved in these projects. Sometimes the same people worked on multiple projects. Even when they didn't, there were often communications between these entities, but sometimes, due to security concerns, there wasn't. Much of this communication was informal, so in retrospect it is hard to tell who knew what. There can be surprises both directions. 3) Communications technologies often emerge naturally from innovations in other fields, technical advances, etc., so while many similar communications technologies have a shared intellectual heritage, it is also not that unusual for totally independent efforts to arrive at roughly the same point. Radio is a classic example. I think that, in consideration of these facts, we can reach two conclusions: A) It seems likely that some of the people involved in ARPANET were familiar with survivable C2 research and applied those ideas to their efforts. After all, lots of people and lots of organizations worked on these programs, and some of the research was widely distributed within the defense-industrial-academic complex during the '50s and '60s. B) It also seems likely that ARPANET independently arrived at similar endpoints. After all, it had some similar constraints and objectives, and its creators were working with mostly the same underlying technology. These two do not contradict each other. In fact, I think it is by far most likely that both are true in the cases of different individual people and different individual aspects of the design. That's just how these things are. Before considering The Internet specifically, let's consider a couple of similar situations in the development of technology: 1) People sometimes do a great deal of hand-wringing over the assignment of labels like "the first programmable computer." I have always been very wary of giving these sorts of titles without a fair number of weasel words. Consider, for example, the ENIAC, usually called the "first programmable computer." And yet, there is a compelling argument that a number of the substantial design elements of ENIAC, including its programmability, are derived from work done for an earlier codebreaking machine called Colossus. This connection remained unknown for many years because of the secrecy surrounding Colossus... a level of secrecy that means that, while a number of people who worked on Colossus and later worked on ENIAC almost certainly carried over ideas, they wouldn't have admitted to having done so as Colossus was officially unknown to the ENIAC project. The particular climate of wartime and military technological development means that ideas often move around in subtle ways, and knowledge of where an idea came from is intentionally obscured. The history of military technology can be a very difficult field for this reason. 2) Almost at the opposite ends of the spectrum, information often flows very freely in academic and industrial laboratory environments, and so ideas spread without clear documentation. I am reminded of a piece I wrote years ago, on the fact that several early internet protocols use a similar set of three-digit status codes in similar ways (HTTP, for example). Oddly enough, these pseudo-standard status codes appear almost simultaneously in RJE and FTP, but neither mentions the other. Over time I have been lucky enough to get in touch with several of the authors of both RFCs, and while none of them can recall the origin of the codes, they agree with my general theory: the two separate groups, both at MIT, had just shared notes during the development of the protocols and one of them informally adopted the status code scheme from the other. People talk to each other, and ideas often move between projects without formal documentation. So, with those two examples of subtle cross-project influence in mind, can we say anything suggestive about ARPANET? Well, there are certain suggestive details. For example, by the time ARPANET's first IMPs were built, at least one researcher (Howard Frank) was engaged in ARPANET research who had previously consulted on survivable C2 networks. But the ARPANET project had already set certain design details like packet-switching by that point... which raises the question of if packet-switching is even the important part. Howard Frank wasn't working on packet switching itself, he was working on performance and reliability modeling of topologies for packet switching, an area where military C2 research was probably generally ahead of ARPANET research at that time. So, if we take the face-value assumption that aspects of ARPANET topology research were probably based on survivable C2 research, does that mean that ARPANET was "created to survive a nuclear strike?" Or did it merely end up that way? It ends up coming down to splitting hairs about what "created" means, an exercise that sort of ignores the fact that technological developments always combine established ideas and new ones. ARPANET was not built for military C2. It was used for military C2 later on, but during the early days of ARPANET the military had more wide-area networking initiatives than you could shake a stick at and ARPANET was not one of the ones contracted for C2 purposes. Was ARPANET designed for nuclear survivability? The most obvious answer is "no," because the early topology of ARPANET lacked the level of redundancy in its topology that actual survivable networks of the era had. But, this seems to have been more a consequence of funding and resource availability than intentions, because ARPANET researchers had done plenty of work on performance and reliability, using basically the same methods as used for survivable networks. So maybe, at the end of the day, the "best" answer to this is sort of a boring one: meh. Nuclear survivability was obviously not a goal of ARPANET because ARPANET did not build out a survivable network. That said, ARPANET incorporated most of the technical ideas from survivable networks of the era. It is a virtual certainty that ARPANET got some of those ideas from earlier and simultaneous research into survivable networks, but it is also a virtual certainty that ARPANET arrived at some of them independently. If you consider "packet switching" to be the main technical advancement of ARPANET, it's probably not an idea that ARPANET got from survivable C2 research, because the historical record looks pretty confident that multiple people independently arrived at packet switching. That ought not to be surprising to anyone, because packet switching is a fairly direct evolution of practices established in radio and telegraph networks almost fifty years before. But, I also think it's an unnecessarily restrictive view of ARPANET's technical contributions, and other aspects of ARPANET like routing policy were definitely influenced by survivable communications research and, to some extent, directly based on survivability work. What all this means about why ARPANET was "created" or what it was "designed" for is strictly a matter of how you interpret those words. Yes, articles and books and etc. should not repeat the claim that "the Internet was created to survive a nuclear strike," because the truth and falsity of that statement requires a lengthy and nuanced explanation. When we express history as simple facts we should try to stick to the ones that are, like, 90% true, instead of like 50% true. But "facts" about history are rarely 100% true, we're just not that lucky. It all happened a long time ago, there were a lot of people involved, different people were doing different things, it's a tangled mess of motivations and influences. That's why we study it. Thanks for coming to my TED talk. Postscript: Also, packet-switching is not at all intrinsic to survivable networks, although certainly survivability lead to a lot of advancements in packet-switching. But there were also a lot of circuit-switched survivable networks, and for a good span of the Cold War, I would say that circuit-switching outnumbered packet-switching for hardened C2. You'll notice that AT&T, the military's #1 choice for hardened communications, was firmly not on the packet switching side of things. But the military also contracted C2 projects to Western Union, who were basically arriving at modern packet switching by their own route (automated telegraph routing). This schism, between packet-switching and circuit-switching, remains a critical part of the data communications story today.