Who Runs the Internet?


 IEEE Spectrum Magazine, March 2002


When I started my career at Bell Labs in the early 1960s, I was an insignificant worker in the vast empire known as the Bell System.  Sometimes I felt like a mechanic in the engine room deep in the bowels of a stately cruise ship.  Somewhere far above me on the bridge of the ship the captain and officers had plotted the course across the boundless seas.  Occasionally, over the din of the machines that surrounded me, I could hear the parties of the happy passengers – the telephone-using public – on the levels above me.  However, my only concern was the health of those engines.  Did they need more oil?  Could they be tuned to run faster or more efficiently?

 I would often see pictures of the smiling company executives in the newspapers.  Everything was under control; the future was planned and assured.  There were shelves of books of Bell System references that detailed every feature of the existing network, and there were scores of reports charting its future evolution.  At that time the future transmission system being developed was the millimeter waveguide, a 5 cm diameter pipe that would carry all the capacity needed well into the next century.  That was a lot of capacity, because everyone would need a 6 MHz analog channel for their Picturephone.  It was all preordained.

 How the world has changed!  A few years ago I got a surprise call from the retired executive who had headed the engineering of the network in that bygone era.  In my mind he had been the keeper of the charts that had plotted the course of that grand ocean liner.  He said that he wanted to find out what was going on in telecommunications today.  Could I meet him for lunch?

At the restaurant he had hardly been seated before he burst out with the question.  “Who runs the Internet?” he demanded.

I was taken aback.  “No one runs the Internet,” I replied, shrugging my shoulders helplessly.

 “That can’t be! “ he shouted angrily, pounding the table for emphasis.

 I remember that conversation occasionally today.  It’s a good question.  Who does run the Internet?  Decades ago we trusted the captain to keep the ship on course towards its proper destination.  But who is steering today?

 In the early days of telephony, back at the turn of the last century, there were competing telephone companies.  Since these companies were not interconnected, it was sometimes necessary to have several telephones to be able to call different people in the same city.  It is easy to see why it was thought that telephony had the characteristics of a natural monopoly and why the Bell System eventually became a regulated monopoly.  Someone had to bring the network together.

The monopoly ended in 1983 with the consent decree that broke up the Bell System, but concern about the integrity of the network led to the creation of Bellcore as a keeper of network standards in the new competitive world.  After all, there had to be some central place to worry about the evolution of the network.

 Today there is no such central place.  For better or for worse, the network steers itself. Assorted standards bodies issue standards that may or may not become fashionable. Manufacturers twist and turn between proprietary systems that may offer competitive advantage and standards-based, me-too approaches.  Service providers ponder which interconnection and peering arrangements they should accept.  Guiding the whole thing in an abstract way are market forces and Metcalfe’s Law, which says that the value of a network grows as the square of the number of users.  In other words, to get maximum value connect with everyone else and be like everyone else.

 I often wonder about how various networks pull together naturally -- fireflies that light in synchrony or generators that pull into a power grid.  When the first digital transmission systems were designed there was a question about how they would be synchronized.  The presumption was that there would be a master clock that ran the network.  However, there was a research suggestion that if every clock in the network adjusted itself to the average of every clock that it saw, that the network would pull into synchrony.  Interesting, but as you might expect, no one was willing to trust that democracy of clocks, and the final design used a master clock.

 Today the Internet pulls in around the Internet protocol, IP, which has been relatively stable for the last thirty years.  It’s like a strange attractor in chaos theory.  Applications may come and go, transmission media may come and go, but IP remains as the stabilizing force at the core of the network.  Only by speaking IP can you be like everyone else.

 So who does run the Internet?  The answer is no one and everyone.  Does this really work?  Maybe it does.  It’s still an interesting question.

 Robert W. Lucky