Minasi Talks ... the Internet

Sometimes it’s useful to get back to basics, and what’s more basic in the Microsoft networking world than the Transmission Control Protocol/Internet Protocol (TCP/IP)? Many people know that a branch of the US military, the Defense Advanced Research Projects Agency (DARPA), funded most of the protocol research that make the Web and Internet email possible. But you might not know why the military offered that funding. The answer lies in the fact that there are 435 members of the House of Representatives, all of whom get to vote on the military budget. The generals are well aware of that number, particularly because most of those representatives come from districts containing at least one company that make computing-related products.

I stumbled upon these facts while computer consulting in the Washington DC area in the late 1970s. Doing any computer-related military work was always a nightmarish experience. Imagine transferring data from a Data General Nova to a Harris H1000. Both minicomputer systems used different processors, computing languages, and OSs; about all they had in common was the ability to read and write a standard sort of magnetic tape, and believe it or not, transferring data by writing it out to one tape drive attached to the Data General, walking the tape over to a nearby drive attached to the Harris, and reading it into the Harris was the standard “interoperability tool.” (We called it the “Sneakernet.”) Standardizing on one vendor wasn’t a possibility, and so the question was always, “How do we get data from vendor A to vendor B?”

Those problems are almost entirely gone in today’s IT world. The bulk of the thanks for that must go to the ubiquity of the Internet. In the 1970s and 1980s, the standard answer to “How do I get data from the IBM mainframe to the DEC PDP-8?” was to buy a special “IBM-to-DEC gateway network” that usually did the job by speaking the network protocols of both vendors but solved only that one problem and was useful if IBM and DEC were the only vendors your organization worked with. But what if the military worked with 50 different types of computer systems? They’d have to buy 50 x 49 /2 or 1,225 different gateway products to enable communication!

The answer was to build a networking system that didn’t speak IBM, or DEC, or Data General, or any other known networking language. Instead, DARPA set up a new set of protocols not married to any particular vendor, OS, or networking hardware: the now-familiar Internet protocols with names like TCP/IP (the basic communication language) and SMTP, FTP and HTTP (a set of high-level command languages for sending email, transferring files for storage or display), and then said to IBM, DEC, and the rest of the computing world, “Please develop software of your own to support these vendor-independent protocols.” It was sort of an “If you build it, they will come” approach to interoperability.

Of course, others had tried this before. They had built it … but very few came. You know that seven-layer Open Systems Interconnection (OSI) model of networks that you’ve probably had forced down your throat at some network class or seminar? Originally, it wasn’t just a model. It described the interrelation of a bunch of actual working protocols being created at the time by an organization once called the CCITT and now the ITU-T. Most of those protocols never really caught on, and so it’s only the model that we remember. So why did the DARPA protocols get vendor support? The answer is simple: DARPA didn’t really say “please.” If you were a computing vendor and wanted to sell your wares to the military (and soon thereafter, the civilian part of the US federal government, as well), you had to support the DARPA protocols. Frog-marching all of those vendors, as well as academics and researchers around the world, onto one network created the critical mass necessary to establish a world-wide network based on a set of protocols that soon didn’t need any mandates to cause it to grow and change. Soon thereafter, public pressure and commercial need caused the US government to open DARPA’s network, now generally known as “the Internet,” to everyone in the late 1980s. And, well, you know the rest.