Summary

  • A November 1996 traceroute reached a telecom host in Addis Ababa; a December operator report described a 256 kb/s TCP/IP gateway through SprintLink; Reuters reported the public service in January 1997.
  • Those dates mark route evidence, network design and customer availability, while the transition also placed the international gateway and its tariffs under the state operator.

Three meanings of “first”

Internet histories often need one date for a process that leaves several kinds of evidence. Ethiopia’s 1996–97 transition is unusually clear because the surviving record separates them. On 28 November, a traceroute reached a telecommunications host in Addis Ababa. On 25 December, an operator described a full TCP/IP connection through SprintLink in Washington. On 3 January, Reuters reported that a domestic Internet service had been introduced for customers. A route, a specified gateway and a service announcement are related milestones; none is a substitute for the others.

The distinction matters because Ethiopia already had computer networks before the full-IP gateway. What was new in 1996 was not the first electronic communication, nor even the first way to exchange messages across borders. The change was a state telecommunications operator’s move to a national gateway that carried TCP/IP traffic to the global Internet.

Email before the national gateway

A 1994 report by the Addis Ababa University Networking Committee describes an effort that began in 1992 with volunteers in the United States, university staff and the United Nations Economic Commission for Africa’s PADIS service. The original proposal imagined a UUCP link. The team instead adopted FidoNet-based messaging after assessing the university’s equipment, staffing and communications costs. By May 1993, a few campus computers could exchange outside email through a dial-up connection to PADIS.

The committee’s April 1994 report says 18 university departments and institutions, plus Alemaya University and the medical colleges at Gondar and Jimma, had email links. It still called an upgrade to the Internet a future need.

Lishan Adam’s May 1995 update likewise says the Addis hub was “still” running FidoNet. He estimated 1,700 users and 285 installed sites, and hoped the country would move to full TCP/IP. Those are a participant’s estimates, not a national census. They nevertheless show that the pre-Internet landscape was not empty: it was a working, locally administered message network with users, calls, equipment and organizational relationships. PADISNet also linked Ethiopian users to international email through store-and-forward arrangements. The ITU’s 2002 case study later summarized PADISNet as connecting daily by direct dial to GreenNet’s London gateway. Email connectivity was meaningful access, but it did not provide the same always-on, routable IP service as the later gateway.

A route appears, then the network is specified

The first of the three 1996 records is a traceroute posted on 28 November. From a machine at Imperial College in London, Olivier Crepin-Leblond traced a path to mail.telecom.net.et. The name resolved to addisababa-server1.telecom.net.et, address 196.27.22.41; the Addis host appeared as hop 16. The reported round-trip times were around 690 milliseconds. It is a striking artifact: a packet from a London computer could reach a named Ethiopian telecom host across a path that traversed several European and US networks. But the traceroute does not establish how many Ethiopian users could connect, whether the service was publicly sold, or the date on which a general customer could first use it. It is evidence of a route to one host.

The operator’s 25 December response to the Network Startup Resource Center answers a different question. Abebe Gashaw identified the connection type as TCP/IP and said the link to the global Internet ran through SprintLink at Washington, DC. He listed 256 kb/s of bandwidth and described a Cisco 7500 gateway router connected to Sprint’s gateway, with domestic routers at various locations to connect through the central router. The equipment list included Cisco 7505 or 2511 routers, Sun Sparc servers and a pool of 64 Microcom modems. The account is a technical description from the service operator, not a measurement of each user’s speed.

The record does not specify whether every planned domestic router was already operating.

This was a different architecture from the scattered email arrangements that came before it. A national gateway offered a legible point for international transit, domestic interconnection, user authentication and billing. It could make a full-IP service technically manageable for one operator. It also concentrated the practical access path: international traffic passed through a gateway whose capacity, connection and customer terms were administered by the telecommunications authority and its successor corporation. “Gateway” describes a router, but it also identifies where operational decisions could be made.

The public service had its own date and price

In August 1996, before the route and December specification, Adam wrote to the AFRIK-IT list that the national telecom authority was working with Sprint to link Ethiopia to the Internet. He added that existing startup networks had been asked to close. Adam presented this as a worrying signal for competition. The archive preserves his contemporaneous account, not a legal order or the authority’s version of events; the claim should be read with that limit. It does, however, show that people building the earlier networks understood the new gateway as an institutional change as well as a protocol upgrade.

Reuters’ 3 January 1997 dispatch, preserved in the NSRC archive, marks the customer-facing milestone. It reported that Ethiopia had introduced its first domestic Internet service, with capacity for 5,000 clients. The head of the Ethiopian Internet Bureau described different charges for government bodies, NGOs and individuals. The same report said that prior access from Ethiopia had depended on costly calls to a provider abroad. In December, Gashaw had listed a $25 rate for five hours, but the surviving wording does not make the billing period clear; he also said a 64 kb/s volume service at $3,000 was not immediately available.

These prices show that access was a commercial and administrative offer, not that the service had already reached 5,000 paying users. Capacity is not subscription.

The distinction among the records remains useful: November establishes a path to a host, December documents the operator’s international link and gateway design, and January reports a service offered to domestic customers. The company’s later history page lists Internet service as launched in 1997, while the ITU case study describes ETC’s service as beginning in January. That retrospective language supports the year and month; it does not erase the earlier technical milestones.

What the gateway changed

It would be misleading to narrate the shift as Ethiopia moving from “no Internet” to “the Internet” on a single day. University email, FidoNet and PADISNet had already connected communities to people and information abroad. It would be equally misleading to call those systems equivalent to the TCP/IP service described by Gashaw. The users’ experience, routing model and available applications were different. The reliable historical claim is narrower: the national operator introduced a full-IP international gateway, described a 256 kb/s connection through Washington, and then announced a domestic service with differentiated terms.

That sequence also shows why an operational detail can carry institutional weight. A gateway centralizes a route in a place where an operator can provision capacity, connect domestic points, recognize customers and set prices. Earlier networks depended on smaller groups coordinating modems, telephone calls, mail queues and local expertise. The new arrangement could expand the scope of network use, while narrowing who controlled the international exit. The archive does not by itself tell us how many users moved from email to IP, how quickly domestic routers reached service, or whether the reported closure request was implemented.

Those questions require further evidence.

The first ping was therefore neither a trivial test nor the whole story. It made a network path visible. The operator’s December note showed how that path was attached to a national gateway; the January announcement showed how it became a purchasable service. In the difference among those records lies the most defensible history of Ethiopia’s Internet arrival: not one celebratory instant, but a transition from locally organized email to a routed, priced and centrally operated connection.

Sources: 1996 traceroute; August 1996 network announcement; operator’s December 1996 network description; January 1997 Reuters dispatch; AAU project report. Additional records: NSRC’s Ethiopia archive; Lishan Adam’s May 1995 FidoNet status.