Larry Roberts (December 21, 1937 – December 26, 2018) was an American computer scientist and an early leader in the development of the Internet.

While working as a program manager and later as an office director at the Advanced Research Projects Agency, Roberts and his team helped create the ARPANET, the first large computer network to use packet switching, a method for sending data in small pieces. This method was first developed by British scientist Donald Davies and American engineer Paul Baran. The main designer of ARPANET was Bob Kahn, along with other scientists from Bolt Beranek and Newman (BBN), who worked on special computers called Interface Message Processors (IMPs) and the rules for how they communicated. Roberts asked Leonard Kleinrock to use math to study how well the network could work. In the 1970s, the Advanced Research Projects Agency supported research on ways to connect different computer networks, using ideas first introduced by Louis Pouzin. This research led to the creation of the modern Internet.

After his work at the Advanced Research Projects Agency, Roberts became the chief executive officer of Telenet, a company that operated the first public data network in North America using packet-switching technology.

Early life and education

Lawrence Gilman Roberts, who was called Larry, was born and grew up in Westport, Connecticut. He was the son of Elizabeth (Gilman) and Elliott John Roberts, both of whom had advanced degrees in chemistry. It is said that during his youth, he built a Tesla coil, assembled a television, and created a telephone network using transistors for his parents' Girl Scout camp.

Roberts attended the Massachusetts Institute of Technology (MIT), where he earned his bachelor's degree (1959), master's degree (1960), and Doctor of Philosophy (Ph.D., 1963), all in electrical engineering. Because of his Ph.D. thesis titled "Machine Perception of Three-Dimensional Solids," he is known as the father of computer vision.

Career

After earning his PhD, Roberts continued working at the MIT Lincoln Laboratory. He read the important 1961 paper titled "Intergalactic Computer Network" by J. C. R. Licklider, which sparked his interest in using computer networks for time-sharing.

In a 1964 MIT video, Roberts explained and demonstrated Ivan Sutherland’s early computer graphics program called Sketchpad, which was running on the MIT Lincoln Laboratory TX-2 computer.

In late 1966, Roberts was recruited by Robert Taylor to join the ARPA Information Processing Techniques Office (IPTO) as the program manager for the ARPANET, even though he was initially hesitant about the administrative role. In February 1967, Roberts met Paul Baran but did not discuss networks at that time. He asked Frank Westervelt to explore design questions for the network. Roberts proposed that all host computers connect directly to one another. However, Taylor and Wesley Clark disagreed with this plan. Clark suggested using dedicated computers to create a message-switching network, later called Interface Message Processors (IMPs).

At the Symposium on Operating System Principles (SOSP) that year, Roberts presented a plan based on Clark’s message-switching idea. There, he met Roger Scantlebury, a member of Donald Davies’s team at the National Physical Laboratory in the United Kingdom. Scantlebury shared research on packet switching and suggested using it for the ARPANET. Roberts applied Davies’s packet-switching concepts to the ARPANET and sought input from Paul Baran.

Roberts’ plan for the ARPANET was the first wide-area packet-switching network with distributed control, similar to Donald Davies’s 1965 design. ARPA issued a request for quotation (RFQ) to build the system, which was awarded to Bolt, Beranek and Newman (BBN). Key aspects of the network, including routing, flow control, software design, and network control, were developed by the BBN IMP team, which included Bob Kahn. Roberts oversaw the implementation and hired Leonard Kleinrock in 1968 to model the performance of the packet-switched network mathematically. He also hired Howard Frank to advise on the network’s topological design. Frank suggested ways to improve network efficiency and reduce costs.

When Robert Taylor was sent to Vietnam in 1969 and later resigned, Roberts became director of the IPTO. He hired Barry Wessler to manage the development of the ARPANET’s host-to-host protocol.

Roberts became a strong supporter of packet switching. In 1970, he proposed connecting ARPANET with Donald Davies’s network at the National Physical Laboratory via a satellite link. This idea was not possible at first, but in 1971, Peter Kirstein agreed to connect his research group at University College London (UCL) instead. This connection created the first international network for sharing resources. In 1973, Roberts predicted that satellite links could be used by multiple satellite earth stations to share data, a plan later implemented by Bob Kahn, leading to the creation of SATNET.

The Purdy Polynomial hash algorithm was developed for the ARPANET in 1971 at Roberts’ request to protect passwords.

In the early 1970s, Roberts approached AT&T about taking over the ARPANET to offer a public packet-switched service, but they refused.

In July 1972, Roberts adapted the READMAIL program, which displayed recent messages on a user’s terminal, into a program for TENEX called RD, which allowed users to access individual messages.

In early 1973, Roberts predicted the network would run out of capacity in nine months. However, the time-sharing host computers reached capacity before the network did.

In 1973, Roberts left ARPA to join BBN’s effort to commercialize packet-switching technology through Telenet, the first FCC-licensed public data network in the United States. He was Telenet’s CEO from 1973 to 1980. Roberts participated in efforts to standardize a protocol for packet switching based on virtual circuits shortly before it was finalized. Telenet later adopted the X.25 protocol, used by public data networks in North America and Europe during the mid- to late-1970s. Roberts supported the X.25 approach over the datagram approach in TCP/IP, which he called “oversold” in 1978.

In 1983, Roberts became president of DHL Corporation. At the time, he predicted that bandwidth would decrease due to voice compression technology.

He was CEO of NetExpress, an Asynchronous Transfer Mode (ATM) equipment company, from 1983 to 1993. Roberts was president of ATM Systems from 1993 to 1998. He was chairman and CTO of Caspian Networks but left in early 2004; Caspian Networks closed in late 2006.

As of 2011, Roberts was founder and chairman of Anagran Inc., which focused on IP flow management to improve quality of service for the Internet.

Since September 2012, he has been CEO of Netmax in Redwood City, California.

Packet switching 'paternity dispute'

In the late 1990s, Roberts stated that by the time of the October 1967 Symposium on Operating System Principles (SOSP), he already had the idea of packet switching in mind, even though he had not yet named it or written about it in the paper he presented at the conference. Some sources describe this paper as "vague." He also claimed that his experiment with Thomas Marill in October 1965 was based on packet switching. Their paper, Towards a Cooperative Network of Time-Shared Computers, published the following year, was described as a plan for the ARPANET. Additionally, Roberts began referring to himself as the "Chief Scientist" at ARPA. These claims have been included in writings about the history of the ARPANET and the Internet and are part of the debate over who first developed packet switching.

Roberts' work at ARPA was mainly administrative and managerial. Before the SOSP, his early research focused on expanding the idea of a support graphics processor to the concept of a network using existing telegraphic methods. In his 1982 paper on "Packet Switching Economics" for the L.M. Ericsson prize, he cited Davies' 1967 SOSP paper but did not mention any earlier works. Historians and primary sources credit Baran and Davies with independently creating the concept of digital packet switching used in modern computer networks, including the ARPANET and the Internet.

Personal life

Roberts married and divorced four times. At the time of his death, his partner was physician Tedde Rinker. Roberts died at his California home from a heart attack on December 26, 2018.

Awards and honors

• IEEE Harry H. Goode Memorial Award (1976), "for his work in designing computer-communication systems, helping create a place where research could grow and lead to new advances in computer and satellite communications, developing standard international communication rules, and creating packet switching technology and the networks that followed from this work."
• Member, National Academy of Engineering (1978)
• L.M. Ericsson Prize (1982) in Sweden
• Computer Design Hall of Fame Award (1982)
• IEEE W. Wallace McDowell Award (1990), "for designing packet switching technology and using it in real life through the ARPA network."
• Association for Computing Machinery SIGCOMM Award (1998), for "important contributions and creating advanced technology for computer communication networks."
• IEEE Internet Award (2000), for "early, important work in creating, studying, and showing how packet-switching networks work, the basic technology behind the Internet."
• International Engineering Consortium Fellow Award (2001)
• National Academy of Engineering Charles Stark Draper Prize (2001), "for developing the Internet."
• Principe de Asturias Award 2002 in Spain, "for designing and building a system that changes the world by offering new opportunities for social and scientific progress."
• NEC C&C Award (2005) in Japan, "for helping create the foundation of today’s Internet technology through the design and development of ARPANET and other early computer networks that were part of the first Internet."
• In 2012, Roberts was added to the Internet Hall of Fame by the Internet Society.