Summary

  • AFRINIC’s March bulletin says its support for Douala-IX is intended to improve local exchange and regional interconnection.
  • Internet Society Pulse’s May snapshot records eight members, 4 Gbps of cumulative port capacity, seven route-server users and five members with at least one valid ROA.
  • Those figures do not measure local traffic, latency, loss, availability, cost or a causal effect of AFRINIC’s support.

The useful question is not whether an internet exchange should be welcomed. It is what a public reader can actually verify today. AFRINIC’s March Community Pulse says the registry supported the Douala internet exchange and describes local interconnection as a route to keeping traffic closer to users and improving network performance. That is a support narrative and a reasonable operational hypothesis. It is not, by itself, a report of measured outcomes.

The independent public snapshot is more concrete and narrower. Internet Society Pulse, drawing its IXP view from PeeringDB, listed Douala-IX in May with 4 Gbps of cumulative port capacity and eight members. It showed that seven of the eight used the route server and that five had at least one valid route-origin authorization. It also listed AFRINIC among the service members. These are meaningful attributes: they describe an exchange with declared capacity, a small membership set, a widely used shared routing facility and some visible RPKI coverage.

None of them answers the claim that needs the most care. Port capacity is not observed traffic. Membership is not a count of active bilateral or route-server sessions. A valid ROA is not proof that every exchanged announcement is valid, accepted or carried. Route-server participation is not a measure of local delivery. The snapshot publishes no time series of bytes or packets, no local-versus-transit classification, no latency or loss distribution, no outage window, and no before-and-after comparison.

It therefore cannot establish that traffic was retained locally, that paths became shorter, or that support by AFRINIC caused a change in resilience or cost.

That boundary matters because an IXP can be operationally valuable long before a public observer can quantify its effect. The absence of a public outcome metric is not evidence of failure. It is simply an evidence limit. Turning it into either a success verdict or a criticism would add facts the two sources do not provide.

An outcome receipt would make the support narrative testable without demanding commercially sensitive disclosure. For a declared observation interval, it could identify the participating ASNs or an anonymised cohort, distinguish route-server from bilateral paths, report aggregate exchange volume and a defined local-versus-transit share, and publish latency and loss distributions from stated vantage points. It would also need the RPKI and IRR rule used, the treatment of outages and maintenance, exclusions, revisions and a clear statement of what changed relative to a baseline.

Aggregate results and a reproducible method would be more informative than a promotional total.

The requested discipline is modest: keep the present public record honest. AFRINIC can say it supported a local-exchange initiative; the public tracker can show the exchange’s visible configuration. Readers should not be asked to infer a measured performance result from either one.

Sources