Ethernet Turns 50 Today: A Day to Review the Feats of Robert Metcalfe, Co-Inventor of Ethernet
50 years ago today, on May 22, 1973, Robert Metcalfe wrote a memo to management at Xerox Palo Alto Research Center explaining how Ethernet could work.
Ethernet is a foundational technology supporting the modern-day internet, and anyone who has had to deal with networking is likely familiar with the IEEE 802.3 standard first published in 1980. Perhaps less familiar is the exact history of how Ethernet came to be and how it remains relevant after decades.
Ethernet turns 50 today, so now is the perfect time as any to celebrate its story and the people behind it. Robert “Bob” Metcalfe—American engineer born in Brooklyn, New York, in 1946—is a key figure in the story of how Ethernet came to be. While Metcalfe is most well-known for co-inventing Ethernet, he proclaims to have had more than six careers.
Robert Metcalfe. Image used courtesy of MIT
Here’s a look at the six careers of Robert Metcalfe.
From MIT to ARPANET
Metcalfe started his academic career at the Massachusetts Institute of Technology (MIT) Sloan School of Management, earning two bachelor’s degrees—one in electrical engineering and one in industrial management—in 1969. He followed with a master’s degree in mathematics from Harvard University in 1970 and stayed there for his Ph.D. in computer science/applied mathematics, which he completed in 1973.
In interviews, Metcalfe commented on his preference for the MIT environment, and during his Ph.D. at Harvard, he took a job back at MIT. His task was connecting MIT computers to the U.S. Advanced Research Projects Agency Network (ARPANET), the first public packet-switched computer network.
Metcalfe’s work connecting MIT computers to the ARPANET was used in his Ph.D. thesis. Image used courtesy of MIT
The Honeywell DDP-516-based Interface Message Processor (IMP) was used to connect networks to the ARPANET and featured a teletype terminal interface. In an interview with Voices of Ethernet, Metcalfe recollects that he had an IMP that was numbered six, while the University of California, Los Angeles (UCLA) was in possession of the unit numbered one. UCLA achieved the first message exchange over ARPANET with a user at a terminal typing “i” then “o” and the receiver terminal echoing the letters back. The goal was to use a predecessor to Telnet to log into the receiving terminal, but in the process of spelling “log,” the receiving terminal crashed.
Xerox PARC and AlohaNet
Metcalfe was still in the process of completing his Ph.D. when he accepted a job at Xerox Palo Alto Research Center (PARC) in 1972. Coming in as the “network guy,” he was tasked to figure out a way to connect the computers at Xerox to a laser jet printer.
Metcalfe had several major challenges to overcome: how to connect all the computers at once from a single line to the printer to avoid creating a “rat’s nest” of cabling, how to ensure that their transmissions wouldn’t collide, and how to encode the information being transmitted.
Some of Metcalfe’s initial experiments involved buying a one-mile spool of coax cable and connecting one end to a square signal generator and the other to an oscilloscope to observe the output. In his initial observations, he saw that by the time the square pulse reached the oscilloscope, the signal appeared more like a cresting wave. Additional components were needed on the receiving end to recover the original square wave.
Robert Metcalfe and David Boggs Team Up
During this experiment, another young engineer observed Metcalfe attempting to strip the insulation from the ends of the cable and struggling to do so cleanly. The young engineer offered assistance, having had experience in the cable industry. This engineer was David Boggs, who would play a key role in co-inventing Ethernet with Metcalfe.
David Boggs (June 17, 1950–February 19, 2022) with an Alto Ethernet card. Alto was Xerox’s answer to the personal computer. Image used courtesy of Ethernet History
To solve the problem of using a single line to the printer, a vampire tap was used at the suggestion of colleague David Liddle. The tap would puncture through the insulation of the main coax cable line and make contact with the copper core.
Example of a vampire tap. Image used courtesy of Canisius University
Tapping in a new line would not disrupt the network and could allow a new computer to be added when needed.
Addressing Signal Collisions
The next problem to overcome involved signal collisions. If two computers tried to send a message at once, they would interfere with one another, and neither would successfully transmit. Metcalfe took inspiration from AlohaNet at the University of Hawaii to overcome this problem.
Unlike the setup at Xerox PARC, AlohaNet was radio-based and allowed multiple computer terminals to send messages to a single receiver. However, since these terminals were not connected to each other, they had no way to coordinate to avoid collisions. In the AlohaNet setup, each terminal was permitted to transmit a message whenever it had data available, adding some degree of randomness that helped minimize collisions. If a terminal received an acknowledgment, it would know that the message was received. If it didn’t, it could try re-transmitting.
Metcalfe found that if a collision occurred and a terminal did not receive an acknowledgment of its message, then it would retransmit after a random short duration. With this scheme, two terminals would not try re-transmitting simultaneously and incur another message collision.
Ethernet Is Born
Finally, Manchester encoding was utilized for its simplicity and independence from a clock signal. With all these elements in place, Metcalfe, Boggs, and others working together at Xerox PARC were able to implement the first Ethernet-based network.
Metcalfe would go on to define an Ethernet protocol that introduced the concept of source and destination addressing, as well as a Cyclic Redundancy Check. In 1980, the first open standard for Ethernet was published as IEEE 802. Ethernet became IEEE 802.3 as other standards became defined by the 802 working group.
Metcalfe remained at Xerox PARC until 1979. During this time, he also began teaching distributed computing part-time at Stanford University.
An image of Metcalfe from a lecture on distributed computing in 1978. Image used courtesy of Texas University
After leaving Xerox PARC, Metcalfe founded 3Com, a network equipment company. After 10 years, Metcalfe departed 3Com and transitioned to a career as a journalist, writing as an internet columnist for InfoWorld magazine, a publication focused on information technology.
In 2001, Metcalfe then became a venture capitalist and in 2010 was appointed as the lead for Innovation Initiatives at the University of Texas at Austin’s Cockrell School of Engineering.
As of June 2022, Metcalfe is a full-time research affiliate at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL).
Ethernet in the 2020s
Ethernet today does not look anything like the Ethernet at Xerox PARC. Today’s Ethernet is more of a packet protocol and no longer uses things like random re-transmission or Manchester encoding.
Metcalfe has stated that Ethernet’s endurance over time may be attributed to the improvements in bandwidth with every successive generation, its open standard, backward compatibility, and competition in the market to innovate and improve it.
Some of the greatest influences in tech have laws named after them, and that’s no different for Metcalfe, who is the namesake of Metcalfe’s law. This law states that the value of a telecom network is proportional to the number of connected nodes in a system.
A single computer in a network has no value since it has no one to communicate with on that network. But if you add a second computer, then that network has increased value since now two computers can communicate with each other. If a third is added, now each computer can communicate with two other computers. Metcalfe's law explains that the value increases with every additional computer added.