Skip to content

Jon Postel: The God of the Internet

Abstract

Jon Postel (1943–1998) edited the internet’s Request for Comments series from 1969 until his death, ran the Internet Assigned Numbers Authority (IANA), which handed out IP addresses, protocol numbers, ports and ultimately the top-level domains, and wrote or co-wrote more than 200 RFCs, among them the specifications of IP, ICMP, TCP and SMTP. In 1977 he argued that TCP should be split in two, which produced the TCP/IP design; in 1983 he asked Paul Mockapetris to design the Domain Name System; and in the TCP specification he wrote the internet’s most quoted engineering rule, the robustness principle. He held no formal authority over any of it. The Economist wrote in 1997: “If the Net does have a god, he is probably Jon Postel.” On 28 January 1998 he asked eight of the twelve root name server operators to take their data from his server instead of the government contractor’s, and they did; the US government objected, he reversed it, and the episode hastened the creation of ICANN, incorporated on 30 September 1998. Postel died sixteen days later.

Jon Postel
Jon Postel. Image: Carl Malamud, CC BY 2.0, via Wikimedia Commons.

Van Nuys and UCLA

Jonathan Bruce Postel was born on 6 August 1943 in Altadena, California. He went to Van Nuys High School in Los Angeles, which also counted Vint Cerf and Steve Crocker among its pupils, so that one suburban high school produced three of the people who later wrote the internet’s founding documents. At the University of California, Los Angeles he took a BS (1966) and an MS (1968) in engineering and a PhD in computer science (1974), supervised by Dave Farber.

UCLA was the first node of the ARPANET. Leonard Kleinrock’s Network Measurement Center sent the first message, “LO”, from Boelter Hall on 29 October 1969, and Postel joined the ARPANET project there on 23 December 1969, as the network was growing from a handful of nodes into a system that needed someone to keep records.

The Editor

Crocker had started the Request for Comments in April 1969 as a way for graduate students to share notes on the network’s protocols without claiming the authority to decide anything; the name was chosen to sound tentative. Postel became the series’ editor and stayed in the role for 29 years. He enforced a plain, consistent format, assigned numbers, kept the archive and decided when a document was ready, and the culture of short, practical, openly published specifications that he maintained is still the way internet standards are written (see The Open Standards Process). He also accepted the occasional joke: the April Fool RFCs, such as the 1990 standard for transmitting IP datagrams by carrier pigeon, appeared under his editorship.

In 1977 Postel moved to the Information Sciences Institute (ISI) of the University of Southern California in Marina del Rey, where he later directed its Computer Networks Division. ISI became the institutional home of what was in practice a personal role. He also helped organize Los Nettos, a regional network for the Los Angeles area, and oversaw its operation from ISI.

Splitting TCP

The first version of the Transmission Control Protocol, designed by Cerf and Bob Kahn, combined two jobs: moving packets between networks and providing a reliable stream of data between two programs. In August 1977 Postel wrote an Internet Experiment Note, IEN 2, that stated: “We are screwing up in our design of internet protocols by violating the principle of layering.” TCP, he argued, was trying to be both a host-level end-to-end protocol and an internet packaging and routing protocol, and the two should be separated. The split began with version 3 of TCP in 1978. The routing half became the Internet Protocol, IP; TCP kept the end-to-end reliability; and applications that did not need reliability, such as voice, could use a thin protocol, UDP, directly over IP. The result is the TCP/IP architecture.

Postel edited the definitive specifications published in September 1981: RFC 791 (IP), RFC 792 (ICMP) and RFC 793 (TCP). In August 1982 he published RFC 821, the Simple Mail Transfer Protocol, on which e-mail still runs.

The Keeper of the Lists

A network of independently written programs needs a single register of numbers: which protocol number belongs to TCP (6), which port belongs to mail (25), which organization owns which block of addresses. Postel kept those lists, at first in RFCs titled “Assigned Numbers”, and then as a function that came to be called the Internet Assigned Numbers Authority. For most of its existence IANA was, in effect, Postel and a small staff at ISI working under a US government research contract.

The lists included names. As the ARPANET’s single hosts file grew unmanageable, Postel asked Paul Mockapetris, also at ISI, to design a distributed naming system, and Mockapetris’s Domain Name System was published in November 1983 as RFCs 882 and 883 (see The Domain Name System). In October 1984 Postel and Joyce Reynolds published RFC 920, which set out the first top-level domains (.arpa, .com, .edu, .gov, .mil and .org) plus two-letter country codes taken from the ISO 3166 list (see Top-Level Domains). Ten years later RFC 1591 (1994) stated his policy for country domains: IANA was “not in the business of deciding what is and what is not a country,” so it would follow the ISO list and delegate each code to a responsible manager.

The Robustness Principle

Postel’s best-known sentence is in RFC 793: “be conservative in what you do, be liberal in what you accept from others.” Known as the robustness principle or Postel’s Law, it told implementers to send strictly correct data but to tolerate small deviations in what they received. It let thousands of independently written and slightly incompatible implementations work together on one network. Security engineers later argued that tolerant parsers accumulate ambiguities attackers can exploit, and that strict rejection is safer (see Secure by Design); in 2023 the IETF published RFC 9413, which recommends maintaining protocols actively rather than relying on tolerance.

Cerf remembered him as “bearded and sandaled… our resident hippie-patriarch at UCLA”, a private person who could be “stubborn beyond all expectation” in an engineering argument. The Economist’s 1997 remark about a god of the Net summed up a position no one had appointed him to.

28 January 1998

By the mid-1990s the domain registry had become a business. Network Solutions, a government contractor, ran registrations in .com, .net and .org and since 1995 had charged for them, and it operated the “A” root server from which the other root servers copied the master list of top-level domains. Postel and the Internet Society had tried to set up an international body to add new top-level domains and oversee the system; the US government, meanwhile, prepared its own plan.

On 28 January 1998 Postel e-mailed the operators of the root name servers and asked them to take the root zone from a server at IANA instead of from Network Solutions’ server. Eight of the twelve did so. The data was identical and users noticed nothing, but for a few days the internet’s naming root followed a researcher’s e-mail rather than the government’s contract. Postel called it a test. The Clinton administration’s internet adviser, Ira Magaziner, reportedly told him he would “never work on the Internet again” if he did not undo it, and Postel restored the original arrangement. Days later the Commerce Department published its “Green Paper” on the future of the domain system, which led in June to a policy for a private, non-profit corporation to take over the functions Postel performed.

That corporation, the Internet Corporation for Assigned Names and Numbers (ICANN), was incorporated on 30 September 1998, with Postel expected to be its chief technology officer. He never took the post. He died on 16 October 1998, aged 55, of complications after heart-valve surgery.

Cerf wrote his obituary as an RFC, RFC 2468, “I Remember IANA”: Postel “had been our rock, the foundation on which our every web search and email was built.” The Internet Society created the Jonathan B. Postel Service Award in 1999 and gave the first one to Postel, posthumously. He was inducted into the Internet Hall of Fame in 2012.

Dead End

Postel’s death ended the internet’s experiment in running critical infrastructure on personal trust. For thirty years the registries worked because one person, known and trusted throughout a small research community, kept them carefully and without visible self-interest. The model had no succession plan, no formal accountability and no legal standing, and the January 1998 episode showed that it also had no limit on what that one person could do.

What replaced it was ICANN: a multi-stakeholder corporation with a board, staff, advisory committees and governments in an advisory role, operating the IANA functions under a contract with the US Department of Commerce until October 2016, when that contract expired and oversight passed to the global multi-stakeholder community. It has been criticized for slowness, cost and politics throughout its existence, and it does not depend on any one person’s health. Open-source projects run by a “benevolent dictator for life”, such as Linux, still operate on Postel’s model and face the same unanswered question about succession.

📚 Sources