Summary

  • A later Angolan registry presentation dates a formal university connection to 28 August 1996, but a contemporary November report says a local Luanda problem had left the reported route without connectivity.
  • EBONet said it tested Internet on 29 October through Angola Telecom, yet in December it was still seeking a licence to begin service. “Connected,” “working,” “tested” and “authorized” described different states.

The network before a public Internet service

Angola was not waiting in silence for one cable to arrive. By 1996, email and information services already linked local organizations to outside networks. But that activity did not automatically provide interactive Internet access.

In February 1996, an Angonet system operator told the Network Startup Resource Center that no organization in Angola then offered “full Internet connectivity.” The message named two active initiatives. Angonet served nongovernmental organizations with email and information services; RIDSANG was a university-coordinated project supported by the United Nations Development Programme and Sweden’s SIDA. The writer described bad telephone lines, costly international calls and irregular electricity as practical obstacles. Angonet hoped to move from FidoNet to UUCP, another store-and-forward system.

That would change how messages moved; it would not, by itself, turn the service into an always-on IP connection.

This distinction was familiar across early networks. A message might leave Luanda in a scheduled exchange, cross an international gateway later and reach an Internet mailbox, even though the sender had no direct route for browsing or other real-time traffic. Angonet’s email system was useful precisely because it worked within those limits. It should not be erased from Angola’s network history, or relabelled as full Internet access after the fact.

A connection with an uncertain life

A 2008 presentation by João Sebastião Teta dates Angola’s formal and technical Internet connection to 28 August 1996. It credits two Portuguese partners: Instituto Superior Técnico and FCCN, the foundation managing Portugal’s academic network, RCCN. In its short chronology, the presentation identifies a university-administered UNINET as a full Internet network in August.

A contemporary field update makes the date harder to treat as a simple opening. Armando Domingues wrote to the NSRC on 8 November 1996 that he had connected Universidade Agostinho Neto to RCCN during a late-August visit, using a 19,200-bps line. He then added that a local problem in Luanda meant “there is no connectivity at all” at the time of writing. The local telephone operator expected to resolve it by the end of November. The same note describes a thin Ethernet, a Cisco router, five 486 PCs, a Linux server and home dial-up modems.

These reports need not cancel one another. One records a formal or technical connection date; the other records a later interruption on the reported local path. Neither establishes continuous uptime. The November writer was describing the Luanda university connection, not performing a national survey. His note shows why an installation date and a usable route are not interchangeable historical facts.

EBONet’s second clock

A commercial service was being assembled at the same time. EBONet’s 5 December update says the company tested its Internet on 29 October and was still working to obtain a licence to start. It identified Angola Telecom as its backbone provider, with a 64-kbps channel through Global One. Its access server had six local dial-up lines, with the company asking for at least ten. EBONet listed its own proposed installation and monthly email-account charges; those are provider-reported prices, not an independent market survey.

In a 2006 history, EBONet founders Silvio Cabral Almada and Haymée Pérez Cogle recall the international output link through Angola Telecom beginning to work on 29 October. They describe a Pentium 100 server, six 33-kbps modems and a dedicated 64-kbps line, and call EBONet Angola’s first commercial Internet server. Their account supplies valuable participant detail, but it is retrospective. The December message’s distinction between a test and a licence request remains part of the record.

The chronology therefore has at least three clocks. A university’s formal connection was later dated to August. A November update recorded an outage affecting that route. EBONet said it tested an international link in October, while still pursuing permission to begin. The 2006 history calls that link an operating commercial server. The sources do not give a contemporaneous commissioning log or the licence itself, so the gap cannot be closed by choosing one of these verbs and discarding the others.

What “first” can safely mean

The date 28 August 1996 can be attributed to Teta’s later account of Angola’s formal and technical Internet connection. The 29 October date can be attributed to EBONet’s report of a successful test and to its founders’ later account of a working output link. November’s outage and December’s pending licence request qualify what those milestones prove. None demonstrates nationwide Internet service, uninterrupted access or a count of users.

Nor do six modem lines tell us how many people connected. The contemporary note says how many lines EBONet had installed and how many it wanted; it provides no session log. A 64-kbps upstream channel shows a particular capacity, not measured throughput or service quality. The February Angonet message reports power, line and price constraints but is one operator’s assessment, not a representative survey of every institution.

The useful history is consequently not one definitive “arrival” date. It is a map of the steps hidden by that label: a university’s international circuit, the telephone network that carried it locally, an outage, a commercial provider’s test, and the licence it said it still needed. The distinction is practical. Technical reach, local operation and permission to sell a service can depend on different institutions, and a dated connection claim alone cannot show that the chain remained intact.

Sources