Summary

  • A March 1997 ping reached one host in Cameroon’s .cm domain from the United Kingdom, but 36% of the probes were lost and the result said nothing by itself about countrywide retail service.
  • A week later INTELCAM was still testing a full IP link via MCI; by 1998, the surviving provider accounts described international connectivity alongside POPs concentrated in Douala and Yaounde.

The measurement was real—and narrow

On 26 March 1997, Olivier M. J. Crepin-Leblond sent a short report to a networking list from London. He could not run a traceroute to Cameroon from his location, he wrote, but a ping to www.camnet.cm, resolving to lom.camnet.cm, had answered. The destination address was 208.148.88.34. Twelve of 19 packets returned: 36 per cent loss. Among the replies, round-trip times ranged from 879 to 956 milliseconds, averaging 923 milliseconds. The surviving note is a measurement, not a declaration that the country had opened a finished Internet service. (NSRC, “First Ping CM,” 26 March 1997)

A ping tests whether a particular host responds to an Internet Control Message Protocol request from a particular point at a particular time. It does not measure how many people can connect, whether they can sustain a session, what a subscription costs, or which towns have a local access point. The 36 per cent loss matters, too: the host was reachable, but the path observed by this correspondent was not a clean, low-delay connection. We cannot infer the cause of the missing packets or extend this small test into a national reliability statistic.

That distinction was visible even before the probe. In October 1996, Jeff Cochrane’s USAID field account described multiple electronic-mail services at the University of Yaounde Polytechnic, each with its own gateway and cost arrangements. A message between two systems in Cameroon could leave the country through one international gateway and return through another. INTELCAM, the national telecommunications company, was discussing a TCP/IP Internet link; local providers wanted to use IP directly. The account captures a transitional communications landscape, not a single, already integrated national network. (NSRC, “Cameroon (10/96),” 20 October 1996)

A responsive host was not a commissioned link

On 7 April 1997, Derek Asoh supplied the missing operational qualification. INTELCAM, he said, was “finalizing the tests” on its full IP Internet link via MCI. Cameroon’s Sustainable Development Networking Programme expected to obtain its own full-IP link when the service became official. The note does not explain how the March ping was routed or which organization operated the responding host. It does show why a single successful probe cannot be treated as the launch certificate for every service layer. (NSRC, “FULL IP now in .CM,” 7 April 1997)

The later record does not yield one uncontested commissioning date. In May 1998, CamNet engineer Jean-Claude Tchoulack wrote that INTELCAM had full Internet connectivity via MCI “since more than one year.” That retrospective places the link’s start before May 1997, but it does not reconcile itself with Asoh’s April 1997 account of tests still being finalized. These are two contemporaneous technical voices describing different moments and perhaps different milestones: a host responding, a carrier link under test, and a service considered official. The archive does not let us collapse them into one day. (NSRC, “Cameroon and the Internet,” 13 May 1998)

The wording also separates backbone reach from retail reach. Tchoulack said POPs—points of presence where customers could connect—were installed only in Douala and Yaounde. Yet he added that a city with a telephone could receive Internet service by subscribing. He quoted an approximate 20,000 CFA francs, or about US$35 at the time. That is a provider engineer’s description, not an independent availability survey. A telephone line somewhere in a city is not proof that every household had a usable line, modem, local number, or affordable plan.

Two cities, several meanings of access

A second note from Tchoulack, dated July 1998, is more explicit about access points and prices. It said Douala and Yaounde were the two cities with “connection room[s]”; elsewhere a POP was not yet available, although users could connect using POPs in major towns. The phrase leaves the service geography unclear. A person might travel to a connection room, dial over an ordinary telephone line to a distant POP, or be describing a more general subscription arrangement. The note does not define those options or their local-call implications, so none should be assumed. (NSRC, “Network Access in the Cameroon,” 17 July 1998)

The price schedule gives the service more texture. Tchoulack listed a 20,000-franc payment for a new subscriber, then monthly plans of 6,000 francs for two hours, 20,000 for ten hours, and 34,000 for twenty hours. A dedicated IP connection cost 1.7 million francs a month, with INTELCAM supplying a modem and router. The May note’s approximate 20,000-franc subscription and the July schedule may refer to different fees or simply reflect imprecise shorthand. The archive does not say which. It does make clear that consumer dial-up and a dedicated IP circuit were different products with very different prices.

A “national connection” can therefore describe several things: one host responding across borders; a carrier’s international IP link; a provider’s POPs; and the places where a customer could actually reach, afford, and use the service. Cameroon’s 1997–98 record contains evidence for all but the last at population scale. It documents an external response, a test process, a retrospective claim of full connectivity, two named POP cities, and a tariff schedule. It does not document nationwide household availability, measured uptime, subscriber totals, or traffic quality across the country.

The point is not to diminish the first ping. It is to preserve what it can prove. A packet crossed enough of the network to get a response from one .cm host. The steps from that reply to an official link, a local point of presence, a purchasable account, and routine use were separate engineering and commercial achievements. When those are kept distinct, the history becomes less like a single “arrival” date and more like the build-out it was.

Sources