Summary

  • AFRINIC’s country export contains both IPv4 prefix rows and a slash_24 field. Counting rows and summing that field answer different questions.
  • South Africa has 2,118 rows and 106,145 listed /24 units; Egypt has 187 rows and 94,343 units. The export does not identify which member holds each prefix.

A row is not a unit

The AFRINIC country-statistics page offers several distinct views: IPv4 in /24 units, IPv6 in /32 units, ASNs, and registered members. Its downloadable ZIP contains a separate CSV for each. That design makes a basic distinction visible, but it also invites a tempting shortcut: count the resource rows and treat the result as the amount of address space, or divide resource rows by member rows and call the ratio a measure of provider capacity.

Neither calculation is supported by the export. In ipv4Data.csv, a row contains a country, a prefix range and a slash_24 value. A prefix can represent many /24 units. The CSV has no organization name, member ID or other stable owner key. memberData.csv lists organization names and countries, but no prefix identifier that can join a member to a resource row. Country is a shared label, not a relationship key.

The distinction changes the picture. In the copy reviewed on 20 October 2026, South Africa appears in 2,118 IPv4 rows. Adding those rows’ slash_24 values gives 106,145 units. Egypt appears in 187 rows, but the same field sums to 94,343 units. South Africa therefore has about 11.3 times as many listed rows, yet only about 1.12 times the listed /24 total. The figures describe the export’s rows and field values. They do not establish unique, non-overlapping address holdings unless overlap and status rules are checked separately.

The member file has 2,629 rows; the IPv4 file has 5,542. Those totals cannot be divided into a defensible “resources per operator” figure. The files do not carry the key needed to connect each resource row to one member, and a country-level match could combine many members with many resource entries. Even a sound member-to-resource join would describe registry records, not the number of operating networks, customers, active routes or deployed capacity.

What the numbers can support

The export supports bounded statements: how many IPv4 rows it contains for a country, how the listed /24 values sum, and how those two measures differ. AFRINIC’s own page labels the country view as issued /24s, while its resource-management guidance explains that resources issued to members are recorded in WHOIS. That is a registry record surface. It is not a census of service providers or a map of where traffic is carried.

Other exclusions matter. AFRINIC distinguishes its delegated statistics, which show allocated or assigned resources, from extended statistics, which also include available, reserved and legacy resources. A reader should not silently substitute a country-export total for the extended inventory, a routing-security check or a utilization measure. Each answers a different question.

The practical risk is a denominator error. A policymaker comparing country totals could mistake more records for more address capacity. A researcher dividing those rows by the member table could create a ratio with no operator-level meaning. A network planner could read registry location as proof of deployed infrastructure. None of these mistakes requires bad data; they arise when a valid field is asked to answer a different question.

AFRINIC can make this surface easier to use by publishing a dated data dictionary beside the download: what one row represents, which resource states are included, how /24 and /32 values are calculated, and whether rows can overlap. A separate aggregate table could expose resource units and member counts by country without implying that the two are joinable at operator level. If an operator-level analysis is a legitimate public need, it requires a documented, privacy-safe method and a clear statement of what the resulting measure does—and does not—represent.

The narrow finding is useful precisely because it is limited. The country export reports registry rows, resource-unit values and a separate membership file. It does not disclose a member-to-prefix join. Until that relationship and the export’s scope are made explicit, record counts can describe the shape of AFRINIC’s file; they cannot count operators or measure the capacity available to a country’s users.

Sources