History of the Internet

Pål Spilling and the Cable That Made Norway Both Reachable and Vulnerable

Norway’s first Internet connection was a door to a global research community—and, for a time, a single physical point at which that door could be shut. Pål Spilling’s career shows why a network’s topology is also a map of authority: it decides who can connect, who absorbs failure, and who may act when speed matters more than consensus.

Louis Pouzin and the Network That Refused to Remember the Conversation

CYCLADES asked what a packet network should be trusted to know. Louis Pouzin's answer was deliberately austere: let the switching system move independent datagrams, and make the computers at the edges responsible for turning them into a reliable conversation. That division of labour became more durable than the French network that demonstrated it.

George Sadowsky and the Workshops Designed to Outgrow Their Teachers

The Internet Society's network-training workshops were built around a demanding test of technical assistance: would the people in the classroom still need the visitors when the next router failed? George Sadowsky's contribution was to make the answer less dependent on equipment and more dependent on whether operational judgment, teaching capacity and trusted peers stayed behind.

Peter Kirstein and the Transatlantic Node That Was More Than a Terminal

When ARPANET reached London in 1973, the scarce asset was not simply a line across the Atlantic. Peter Kirstein's achievement was to turn UCL from a distant customer of American machines into a place where British researchers could attach computers, implement protocols, operate gateways and acquire a voice in how networks joined.

Dai Davies and the Buyer Europe Did Not Yet Have

In 1993, Europe’s national research networks did not mainly lack ambition. They lacked a counterparty. Someone had to turn many countries’ requirements into one specification, take that specification to a guarded telecommunications market, award contracts, reject deficient delivery and return for the next network with stronger bargaining power. Dai Davies’s consequential achievement was helping to make that someone exist.

Klaas Wierenga and the Credential That Stayed Home

A student from Delft opens a laptop in Cape Town. The local university can offer a network, but it cannot vouch for the visitor’s account. Delft can vouch for the account, but it does not control the Cape Town network. Klaas Wierenga’s decisive contribution to eduroam was to make that division of authority the architecture, rather than a problem to be concealed behind a universal login service.

Jan Gruntorád and the Link That Had to Become Two Networks

At the start of 1992, an international circuit from Prague to Linz was strengthened and divided. One channel continued to carry EARN traffic; the other was assigned to the Internet. In that physical line becoming two operational networks lies a more useful history of Jan Gruntorád than any story about a single day when a country suddenly “went online.”

Ben Segal and the TCP/IP Pilot That Could Not Leave CERN

By March 1988, TCP/IP served 355 machines at CERN, from an IBM mainframe and a Cray to VAX systems, workstations, PCs and Macintoshes. Yet none of those machines could use TCP/IP to cross the laboratory’s external boundary. Ben Segal’s pilot had solved heterogeneous connectivity before CERN had decided what that success was allowed to connect to.

RFC 2057: the 1996 argument over who controlled the audience

In 1996, sending an email to a mailing list could hand its onward distribution to another operator's program. Scott Bradner's RFC examined the distance between controlling that first transmission and establishing the identity and age of everyone it eventually reached.

Rob Blokzijl and the Chair That Could Coordinate but Not Command

The most revealing sentence in RIPE’s first terms of reference is not the one that declares a European ambition. It is the sentence that withholds power. RIPE was not a network provider, and the networks cooperating through it remained under the executive authority of their own organisations. Rob Blokzijl would chair that arrangement for twenty-five years. His office mattered because it could make independent operators act together without pretending to own the machinery they operated.

Qiheng Hu and the 64K Link That Needed Authority at Both Ends

By the end of 1993, the network was technically ready. Its domestic backbone joined three academic systems, its team had chosen TCP/IP, and the international circuit had been provisioned. Yet packets still did not cross. Qiheng Hu’s part in China’s full-function Internet connection began where a diagram of routers and a 64K line stopped being an adequate explanation.

RFC 2056 and the Fork in a Z39.50 URL

In 1996, RFC 2056 gave two URL schemes to Z39.50 because one locator could conceal two very different kinds of work. One scheme entered a continuing, stateful inquiry; the other attempted a tightly bounded retrieval of a server-identified item. The distinction is a small piece of Internet standards history, but it captures a larger design problem: how much of an application's operating intention should be visible before a protocol exchange begins?

Jun Murai and the Address Space That Became a Trust

The number is oddly exact: 14,680,064 IPv4 addresses. They were neither a fresh registry allocation nor abandoned space waiting to be reclaimed. They were seven-eighths of an old Class A block, held through three decades of changing Internet institutions, and in 2020 Jun Murai chose to move them into a trust.

Walk Could Not Be Repeated: RFC 2041’s Contract Between Trace and Experiment

The Walk Could Not Be Repeated: RFC 2041’s Contract Between Trace and Experiment

A wireless failure observed along one route could be taken back to a laboratory, but only after researchers decided what the record meant and which effects a wired testbed could reproduce. RFC 2041 made those decisions unusually visible.

Kilnam Chon and the Router Asia Could Not Buy

On 15 May 1982, two computers in South Korea began speaking across a 1,200 bps leased line. The connection was modest; the missing box beside it was more consequential. With no obtainable IP router, Kilnam Chon’s group treated published protocol documents as equipment they could build.

Zero Handle Opened a Door. It Did Not Grant Entry

The Zero Handle Opened a Door. It Did Not Grant Entry

In 1996, RFC 2054 tried to remove two conversations that stood between an NFS client and its first useful file reference. Its answer was not a new permission system. It was a disciplined gamble: try the port most servers already used, try the newer protocol first, and present a reserved empty handle as the starting point. Every shortcut came with a named retreat.

A Cipher Name Was Not Enough: RFC 2040 and the Missing Identity of RC5

RFC 2040 reads less like a manifesto for a cipher than a parts ledger for agreement. Its lasting historical value is the discipline with which it separates a family name from an executable identity: word size, rounds, key length, expanded state, chaining state and the rule used when the message ends.

Kees Neggers and the Backbone Europe Built in Parallel

On 26 September 1992, one link between London and Montpellier completed Ebone’s initially defined backbone. The more revealing number was not its line rate but its age: 390 days since a small group had first met outside Amsterdam to ask whether Europe could operate a shared IP backbone while official policy was still looking elsewhere.

Configuration Row Vanished. RFC 2051 Kept the Session Alive

The Configuration Row Vanished. RFC 2051 Kept the Session Alive

RFC 2051 turned an obscure APPC management model into a durable lesson about evidence: the record of what should exist can disappear while the record of what is already running remains true.

Karen Banks and the Gateway That Did Not Need to Stay Online

A message could wait on a local machine, travel during one brief telephone call and still reach an international public. Karen Banks’s work on GreenNet’s GnFido gateway shows why early access was a chain of queues, operators and partners—not a light that simply switched from offline to online.