The National Inventors Hall of Fame has inducted Bob Metcalfe, the alleged inventor of Ethernet. This perpetuates a myth that Metcalfe has been stoking for 30 years, and it’s wrong. While it’s certainly true that Metcalfe was one of the people at Xerox PARC to co-invent a network called Ethernet in 1973, that network has very, very little to do with the network we call “Ethernet” today.
Metcalfe’s Ethernet was a coaxial cable shared by a number of computers, each of which connected to it through a little radio-like thing called a transceiver through a bundle of wires as thick as a pencil.
The network we call Ethernet today is a box of digital electronics called a switch or a hub that the computer connects to through two pairs of twisted copper wires. On the modern switched Ethernet several computers can communicate at the same time, but on Metcalfe’s system one and only one could transmit at a time.
You can only credit Metcalfe with inventing Ethernet if you expand the meaning of the term “Ethernet” to include all local area networks invented after 1973 and ignore the older ones, like AlohaNet. Metcalfe seems to be encouraging that sort of thing, as he’s recently described the Cable Internet protocol, DOCSIS, as an Ethernet:
“DOCSIS is Ethernet,†he claimed. “It’s HFC [hybrid fiber-coaxial] Ethernet.â€
Bob Metcalfe invented the name “Ethernet”, but he didn’t invent the modern technology that goes by it.
Can we finally get this straight? The guys who had the most to do with creating the network now known as Ethernet are Tim Rock of AT&T Information Systems and Bob Galin of Intel. They were members of an IEEE 802.3 task force on low cost networking formed in 1984 that produced the 1BASE5 standard. The network they invented, once called StarLAN, evolved into 10BASET and was then renamed Ethernet. And I know all of that because I was the vice-chair of that committee. So let’s give credit where it’s due.
Bob Metcalfe is a clever guy with a talent for public relations, but he’s not the father of modern local area networking.
The original coaxial Ethernet was almost unworkable. In an office or lab environment, where computers were being moved around frequently, the cable connections were easily damaged. Any break or degradation of the signal quality on the cable would effectively cut off the computers on that segment of the bus from all the other computers on the bus. Something as simple as poor impedence matching or mis-termination degraded the signal. Communications protocol errors at a single node on the bus brought down the entire network. Troubleshooting a broken network was a time-consuming and frustrating process. Troubleshooting a degraded network was almost impossible.
LANs were not physically practical until the star topology emerged. Star topology made troubleshooting easy. LANs were not logically practical for communications until intelligent hubs, switches, gateways, and routers were invented.
Bob Metcalfe “was in the room” at 3Com when many of the things that made Ethernet practical were developed at his company and elsewhere, but he did not invent them any more than Al Gore invented the internet.
And Metcalfe said that he did invent them a lot more, and Al Gore didn’t.
3Com actually fought against the hub-and-spoke system aggressively. After the StarLAN standard was ratified and that committee’s members went on to draft 10BASE-T, 3Com proposed a single pair system that would have used the bus topology that made co-ax Ethernet such a pathetic system. The committee shot it down and carried the StarLAN system forward, and the rest is history.
Can you provide me more detail on the development of StarLAN and its migration to IEEE 10BaseT? Am writing a history of telecomm industry and would appreciate your first hand knowledge.
[email protected]
Intel noticed in 1983 that its 82586 Ethernet chip wasn’t selling and did some research, coming to the conclusion that the Ethernet system was way too expensive to deploy. They then built a scaled-down Ethernet chip (the 82588) for cable tv coax and twisted pair at 1-2 megabits/sec. NCR and AT&T were on board, and together with Tandem petitioned IEEE 802.3 for a study group on low-cost LANs.
The group considered both coax and twisted-pair approaches and concluded that twisted-pair was the way to go. In 1987, the 1BASE5 standard was ratified, and members of that task force, including Pat Thaler of HP, went on to scale it up to 10 megabits with 10BASE-T, which also added the full-duplex option. Pat was a core member of the 1BASE5 Task Group and Chair of 10BASE-T.
That’s the short version.