Summary

  • LACNIC Blog reports at least 138 forwarding patterns consistent with possible routing loops in measurements of Dominican Republic IPv4 prefixes. Its first classification is 24 intra-AS, 14 inter-AS and 100 unclassified.
  • A second rule treats a case as intra-AS when the last public hop before a private-address cycle and the destination prefix belong to the same ASN. The published count then rises to 83 intra-AS, leaving 41 unclassified; the 59-case shift is derived from the totals, not separately itemised.

Analysis

The headline number is less informative than the boundary around it. Wilson Ramírez’s LACNIC Blog study does not report 138 confirmed router failures. It identifies repeated, periodic hop sequences in traceroute measurements as patterns consistent with possible loops. The distinction matters: a trace can show packets returning through the same visible addresses without establishing why that happened, which equipment was responsible or whether users experienced an outage.

The collection began with 1,476 IPv4 prefixes announced by 197 ASes registered in the Dominican Republic. The team sent ICMP probes to at least 40 addresses in each prefix. Where a prefix produced at least one ICMP Time Exceeded reply, it recorded the replying address and selected another target in that same prefix; 234 targets went on to additional traceroutes. The article names RIPE Atlas probes 60205 (AS6400, Claro Dominicana) and 32174 (AS267705) as vantage points. It also describes ICMP- and TCP-based traces, custom checks through NetKitBox and DNS-checker, and repeat measurements at different times.

The initial classification leaves a substantial evidence boundary. Fourteen cases were assigned across ASes. Twenty-four were assigned within one AS. In the other 100, private addresses appeared inside the repeated cycle or intermediate timeouts made the AS identity unreliable. The alternate rule uses two visible endpoints around that gap: the final public hop before the private-address segment and the destination prefix. If both map to the same ASN, the study counts the case as intra-AS. That produces 83 intra-AS cases and 41 still unclassified.

The arithmetic implies 59 cases changed buckets, while the inter-AS count stayed at 14. It does not mean the private routers were directly identified. A destination prefix’s ASN and the last public hop can support a practical hypothesis about where a cycle sits; they do not expose the ownership or control of each hidden hop. The rule makes the classification more complete by inference, but it also makes the inference itself part of the result.

The study separately labels loop shapes as Ping-Pong, Circuit or Unclassified, and says Ping-Pong is the majority without giving a percentage in the text. Five cited RIPE Atlas measurement IDs remain publicly retrievable and show representative repeated hops. Those examples help a reader inspect the phenomenon; they cannot reproduce the aggregate count or establish how common it is across the country. LACNIC’s blog also notes that contributors’ views do not necessarily represent LACNIC’s institutional position.

For operators, the useful question is therefore not whether 138 is a national failure rate. The post gives no per-AS or per-prefix incidence denominator for that calculation. The actionable issue is whether an escalation labels a loop from a directly observed hop, a time-out-bounded trace or the broader last-public-hop rule. Those are different strengths of evidence, and the published totals show how much the choice can change the apparent location of the problem.

Sources