To kick start the U.S. economy we need more engineering chutzpah. This week Washington is unveiling a fix for the banking industry along with a stimulus package to resuscitate the moribund economy. These are necessary temporary boosts–but for the long haul the economy needs new sources of sustainable growth. Alternative energy and green cars top the list of promising foundation stones for long-term domestic economic growth. Our new president is urging us turn our sights to science. “We will restore science to its rightful place,” President Obama declared in his inaugural address. We love science and believe it is important, but will the next new thing come from scientists or engineers? We vote for the engineers.
In our view Henry Petroski of Duke University has it right in a recent Washington Post article. He makes a compelling case to turn to engineers rather than scientists for breakthrough innovations. Using the steam engine and the airplane as examples, he explains that in the past real inventors were not “hampered” by science. In fact the scientists of the day routinely impeded progress by declaring that the big new things were impossible dreams.
Peter witnessed this phenomenon first hand in the history of data networks. Here is his story.
Packet switching was invented as a theory in the 1960s and became a reality when BBN launched the ARPANET in October of 1969. The ARPANET quietly expanded to 23 nodes by April 1972. Despite the technology’s success it was essentially unknown until the International Conference on Computer Communication in Washington, DC in October of 1972 where the ARPANET was demonstrated. Interest generated at the conference created a big ARPANET traffic spike and exposed several hundred engineers to the technology.
I was one of the attendees who tried every demonstration multiple times. I was so excited about packet switched networks that I dragged my wife to the conference the second day to show her the future. We stayed late into the night. She was amused, and I was hooked.
One year later found me developing a packet switch at the research lab of Northrop Page. Several people in the US government and industry were already planning to build new packet switched networks, and we all realized we needed an interface standard for the new technology. We petitioned standards organizations to define a new computer-network interface.
Our lobbying led to several international working meetings in 1974 hosted by the CCITT Rapporteur on Packet Switching. A Rapporteur’s job is to determine the need for a standard. Good insights were available from the ARPANET, Telenet, French CYCLADES network, and a few research labs. By then our small international contingent of experts was ready to press the case that packet switching was an important technology requiring a new interface standard for public data networks.
I attended the general CCITT meeting where the Rapporteur presented his conclusions and we expected to start formally defining a new standard. The international audience proved cordial but unreceptive. Even at this early stage, issues and forces emerged that remain with us. The end-to-end principle is an example. The US was defining the datagram that expected little in services from the network, while carriers wanted network-based control over all services. The outcome of that meeting was we were told to come back later.
A private conversation with a member of the German delegation shed light on that outcome. He explained that there should not be a standard for packet switching at that time. Here was his logic: “You Americans have an interesting experiment in the ARPANET. But there is no formal mathematics underlying packet switching technology. You can’t mathematically prove that sending individual datagrams via alternate paths works. If there is no math, then there is no science–and if there is no science there can be no technology. We do not standardize things that are not a true technology.”
We went back to the US and built the Internet anyway, and the math and science are still catching up with how the technology actually works. We need inventors and engineers to break new ground, and the establishment should not stand in the way.
By the way, the CCITT standard for packet switching is called X.25. The X.25 packet switch market had a brief run and there are still some X.25 services operating. But today, there is no market for X.25 switches except as spare parts to keep the networks running. That pesky datagram called IP (Internet Protocol) won out in the end.
