Larry Roberts was a Networking Legend

Internet old-timers were deeply saddened by the passing of networking pioneer Lawrence G. Roberts on December 26. Larry is the first High Tech Forum contributor to pass away: he wrote a piece on wireless billing plans for us in 2010.

Larry’s contributions to networking are unparalleled and under-appreciated. Not only did he design ARPANET, the proof-of-concept for packet switching, he enabled people all over the globe to connect to ISPs without paying long distance telephone charges through the Telenet packet switched data network.

ARPANET proved that packet switching was not only viable, but the only feasible way for people to use computers at a distance. Larry took this learning to heart by founding Telenet and providing a foundation for both public and private computer networks of all kinds. Larry not only enabled the Internet to be built, he created a technology that will outlast it.

Building ARPANET

ARPANET was built in the late ’60s to allow researchers to share expensive computers located in research labs around the country. It was created at a time when computers cost millions of dollars and telecom was also very pricey. Packet switching was the perfect solution for this problem because it was ideal for scenarios in which data volumes were low in relation to connection time.

While the telephone network is very limited in terms of data volume, it’s always active for the entire duration of a call. It takes a relatively long time to establish a telephone connection that enables a relatively small amount of data to be transferred. Computer networks need faster connections and the ability to transfer information in clumps.

Packet switching accomplishes this by making a high capacity pipe available to multiple users at the same time. As long as most users are not transferring data at the same time, it’s great; and that’s the common scenario.

While packet switching was initially invented by Paul Baran at RAND in the early ’60s, most of his work was classified so the idea had to be recreated by Donald Davies in the UK before Larry could learn about it.

The Legacy of ARPANET

ARPANET lasted 20 years, a good run for a technology advance. It had performance limitations reflecting the limited memory available on the computers of its era. When the typical computer only has dozens of kilobytes of memory, it’s easy to overrun it with incoming data. As computers grew more powerful, networking researchers were able to extend the ARPANET paradigm to make better use of new capabilities.

A lot of that early research was done by people who worked at one time for BBN, the firm Larry selected to build ARPANET. BBN was a strange choice in some ways because its primary field was acoustics. But when nobody has ever built a packet switched network, everyone is equally qualified.

Two BBN engineers – Alex McKenzie and Dave Walden – worked with Louis Pouzin‘s CYCLADES team in France to build the network that was the paradigm for TCP/IP, the Internet’s foundation protocols. CYCLADES engineer Gérard Le Lann was a member of Vint Cerf’s team at Stanford that created the TCP/IP design, which also owed a lot to INWG 96, primarily created by McKenzie.

The beauty of the Internet, of course, is its ability to make use of advances in the speed and reliability of telecommunications networks. Unlike common networks, the Internet is virtual rather than physical. Hence, it makes use of telecommunication networks of various types even though it’s nothing but software, specifications, and agreements in its own right.

The Importance of Telenet

As we explained in our Amicus Brief in the current challenge to the deregulation of Internet access, Larry founded Telenet in 1972, long before the Internet was designed. This history is explained in a paper by Larry and colleagues, The History of Telenet and the Commercialization of Packet Switching in the U.S.

Early online services ran on centralized computers accessed by users spread across the nation and even the world. Few could afford to pay long distance tolls based on connection time, especially when data volume was low. By cutting communication charges to a tenth of long distance rates, Telenet enabled online information services of all kinds viable.

Its first customer for this kind of service was The Source, the founders of which went on to create America Online. AOL joined Barry Shein‘s The World, another user of packet data networks, in opening the fledgling Internet up to ordinary people who weren’t working for universities, pursuing advanced degrees in computer science, or doing government-sponsored research.

While it’s common to assume that the dialup ISPs relied on the telephone network for user connectivity, the role of telecom was limited to connecting local calls and providing high-capacity leased lines for the Internet backbone. In reality, Telenet (and similar firms such as Tymnet) made these businesses work.

Roberts was Right on Internet Policy

Paul Baran’s initial packet switching design was a voice network, essentially a hardened, limited access telephone network. While some of the most interesting uses of the Internet involve voice, such as Tom Evslin’s ITXC and Jeff Pulver’s Vonage, the Internet doesn’t handle voice as well as it could.

This always frustrated Larry because he wanted packet switching to handle all of the world’s communication. Many people still maintain hard line phones even today because they feel they’re more reliable than the Internet. And in some ways they’re right.

Today’s Internet billing plans hide a multitude of complications. While consumers want flat rate pricing, the costs of providing Internet service depend, to a large extent, on usage. Harmonizing diverse uses on a common network is an ongoing research topic, and it probably always will be.

Creating a network that can be all things to all people was a monumental undertaking. Making it work for every user in the most reliable, safe, and economical way is even harder. I happily shared the Amicus Brief with Larry last October that was influenced so heavily by his work on Telenet; and I was glad that it pleased him. His last message to me was a thumb’s up emoji.

There’s never going to be another Larry Roberts. Enjoy this video of a talk he gave at the Computer History Museum courtesy of ISOC.