Summary

  • In 1990, LEARNmail gave Sri Lankan academics an email service built around dial-up UUCP, while a separate connection linked Moratuwa with Stanford.
  • A later participant account put the international call at about US$1 per minute and said mail initially moved once daily; later funding raised the reported exchange frequency.
  • This was not yet a permanent, always-on global Internet connection. A domestic IP wide-area network and LEARN’s full Internet connection came later, on separate timelines.

A network begins as a question of when to call

In the early 1990s, “being connected” could describe several different arrangements. A university might exchange messages through dial-up software, two campuses might reach each other over a local data link, and a research network might still lack an always-on route to the global Internet. Sri Lanka’s early academic network is a useful case because the sources preserve these distinctions—and show how the price of one part of the system shaped the design of another.

A 2016 account by APNIC records Gihan Dias’s recollection of an international UUCP dial-up connection between the University of Moratuwa and Stanford University. Dias described a cost of roughly US$1 per minute, high enough that email was transmitted once a day. With further funding, the service reportedly moved mail two or three times daily. The same account dates LEARN’s establishment to April 1990 and describes it as an academic network conceived before commercial Internet service was available locally.

The number is a participant’s retrospective estimate, not a surviving telephone invoice. The article therefore treats it as evidence of how the organizers understood the constraint, not as a verified tariff for every call. Still, the mechanism is clear in the account: when each minute of international connectivity carried a high marginal cost, users and operators had an incentive to collect messages, exchange them in batches and disconnect.

UUCP moved messages without making a permanent Internet link

UUCP—Unix-to-Unix Copy—was suited to intermittent links. A machine could queue mail and other files, establish a call, transfer what was waiting, then end the session. That made useful communication possible without keeping a long-distance circuit open. It also meant that “email worked” did not imply that a sender and recipient had a live end-to-end session, that messages arrived immediately, or that every campus had direct international access.

APNIC’s history describes the connection as one organized by LAcNet, a group of Sri Lankan academics abroad, with support from the University Grants Commission and CINTEC, the country’s computer and information technology council. The account says this arrangement helped give rise to the Lanka Experimental Academic and Research Network, later known as LEARN. A 1996 report by Abhaya Induruwa to the Computer Society of Sri Lanka retrospectively places LEARNmail among the network’s earliest milestones, dating the first IP-based email service to 1990.

The phrase “IP-based email” in that report should not be mistaken for a claim that the whole country already had a routed IP network or public Internet access. Email service, transport protocol, domestic wide-area connectivity and an international Internet uplink were separate layers. Later institutional histories sometimes compress them into a single “arrival of the Internet” date; the records are more informative when kept apart.

Building locally changed the shape of the problem

The initial international exchange connected Moratuwa to a machine at Stanford. It did not, by itself, join Sri Lankan universities to one another over a domestic backbone. LEARN’s broader purpose was to serve academic and research communication inside the country. The sources describe a domestic IP network linking the University of Moratuwa, the University of Colombo and the Open University through wireless links, with the implementation commonly dated to 1994–95. They differ in how they date commissioning and completion, so this article does not force one exact launch day.

The domestic network changed what had to cross the costly international leg. Once campuses could communicate over a national network, the international connection no longer needed to carry every local exchange. That is an architectural implication of the topology described in the sources, not a measured traffic outcome: no packet counters, route tables or call logs establish how much international traffic was displaced. But the order of development—intermittent international email, then domestic IP connectivity, then a permanent global Internet link—shows that local reach and global reach were built as different capabilities.

A 1996 report by Induruwa says LEARN connected to the global Internet in January 1996, with the first year’s link financed by CINTEC. Other retrospective accounts associate a permanent connection with 1995. Rather than choose one as the sole national “first,” the useful distinction is between the commercial demonstration or initial service, a dedicated connection, and the academic network’s documented online link. They are not necessarily the same milestone.

What the evidence can and cannot show

The available record supports a bounded history: academic organizers used dial-up UUCP to move email between Sri Lanka and the United States; a participant recalled high per-minute expense and daily batching; LEARNmail began around 1990; local IP networking connected several universities in the mid-1990s; and LEARN reached the global Internet on a later, separately financed link. The record does not establish the exact tariff, call duration, user count, message volume, delivery latency, routing configuration, or how much the domestic network reduced international calls.

The distinction matters because a dated connection is not a usage measurement. A 64 kbps link tells us a nominal capacity, not utilization or reliability. A launch announcement does not establish that every intended campus was connected at once. A recollection of daily mail gives a human scale to intermittent networking, but it is not a complete service log.

The durable contribution was not that one dial-up call somehow made Sri Lanka “online.” It was that researchers made communication useful under a narrow operating window, then pursued domestic infrastructure that could carry local exchange on local links. The resulting sequence—queue, call, transfer, disconnect; then campus network; then global uplink—reveals how connectivity was assembled in layers. The country’s Internet history is clearer when those layers remain visible.

Sources

The chronology combines a participant interview, a January 1996 progress report and later institutional or scholarly retrospectives. These establish the broad sequence, not a complete traffic or billing record. The dollar-per-minute estimate and batching frequency are attributed recollections; precise operating dates differ among sources.