Summary

  • LACNIC’s current regional study describes Speedchecker-based paths from twenty origin countries, yet its public report does not state an observation window, current sample matrix, dataset identifier or analysis version.
  • The report points readers to an earlier LACNIC study that explicitly documents a 2022 measurement campaign. That link makes lineage more important, but it does not prove that the current figures reuse the older traces.
  • The publication date, report date and measurement time are different facts. Without their join, an operator cannot tell whether a chart is a fresh baseline, a continuation or a reanalysis.
  • A compact edition manifest could make the evidence locally verifiable without turning LACNIC’s analysis into a central instruction for how networks should route.

Two clocks, one label

Every measurement report carries at least two clocks. One belongs to the publication: when an edition was written, uploaded, revised or announced. The other belongs to the observation: when the probes sent packets and the network produced the paths that were recorded. The first clock governs citation. The second governs interpretation.

LACNIC’s public materials make the publication clock fairly easy to reconstruct. Its technical report entry is dated 15 September 2025 and says it was modified on 8 January 2026. The accompanying English blog post appeared on 12 January 2026 and calls the work the “2025 Traffic Paths Study.” The entry names Elisa Peirano as author. Its Spanish PDF contains thirty-five figures, while the visible version and project fields on the report page are blank. LACNIC technical report LACNIC blog post

The observation clock is not supplied. In the public report and its downloadable PDF, I could not find a start or end date for measurements, a campaign identifier, aggregate counts for probes, origins, destinations or traceroutes, a table of samples by country or country pair, dates for the source snapshots, a code or method version, or a link to a raw dataset. The annex is a sequence of figures, not a machine-readable ledger connecting those figures to one collection run. Spanish report PDF

That absence does not make the study worthless. It changes the question a careful reader can answer. The report can show what its authors found in the material they analysed. It cannot, on the public record alone, let an operator establish when that material described the network.

What the study usefully puts on the table

LACNIC’s blog describes measurements obtained through Speedchecker, with paths originating in twenty countries: seven in Central America, ten in South America and three in the Caribbean. It invites operators to compare the findings with their own data and stresses that adequate coverage matters. That is the right posture for regional measurement. A chart is most valuable when it prompts a local check, not when it claims the authority to replace one.

The report’s conclusions are operationally interesting. It says that, for country pairs involving destinations outside the region, country-level average latency runs from 100.9 to 178.6 milliseconds. For traceroutes whose origin and destination lie in the same subregion, it says between 48 and 71 percent leave that subregion and later return. Those numbers can focus attention on interconnection, transit and the geography of path selection. They are attributed findings from the report; this article has not reproduced them from packet traces or tested whether they describe routing today.

The study also shows why aggregation needs context. A path seen from a measurement platform is not “the route of a country.” It is one observation conditioned by the selected source, destination, access network, resolver and time. Country-pair summaries can reveal recurring shapes, but the mixture underneath matters. A shift in access-network coverage can move an average even if no operator changes policy. A stable average can hide large changes among individual paths. Neither possibility is an objection to aggregation. Both are reasons to publish the sample frame beside it.

The current report excludes Aruba, Belize, Cuba, the Falkland Islands, French Guiana, Guyana, the former Netherlands Antilles and Suriname because it says there were insufficient results. That is a defensible analytical boundary. It must not be converted into a claim that those places lacked connectivity. “Not enough observations for this comparison” is a statement about a dataset, not a statement about a country’s network.

The earlier campaign sharpens the lineage question

In its data section, the current report links an earlier LACNIC study and says that the considerations described there remain valid. The earlier English report is much more explicit about collection. It says measurements were conducted in 2022 through Speedchecker, identifies 5,588 networks in scope and 511 intermediate networks, applies a threshold of fewer than ten results, and says its graphs were prepared from raw CSV files. Earlier study page Earlier English report

The current exclusion list closely resembles the earlier exceptions, though it is not identical. That resemblance is a clue about why lineage should be explicit. It is not evidence that the new report reused the 2022 traces. The present public materials do not establish whether the newer figures reuse, extend or replace that campaign. Any of those workflows could be legitimate. A reanalysis can be valuable. A continued campaign can be more informative than a one-off run. A fresh campaign can support comparison with the past.

The problem is not choosing one of those methods; it is leaving the reader unable to tell which method produced the edition.

A PDF’s upload path or file-creation date cannot fill the gap. Files are regenerated, copied and republished. A report may be written months after collection or updated without recollecting data. Treating the document timestamp as the measurement timestamp would manufacture precision the source does not provide.

A small wording conflict shows what a manifest should fix

One section of the Spanish PDF explains box-plot whiskers as the first and ninetieth percentiles. Later text describes them as the minimum and maximum, while the figure axis labels indicate p1–p90. This is a bounded documentation inconsistency. It does not invalidate the figures, and it would be reckless to infer which calculations ran from prose alone.

It does reveal why the method must be attached to an edition rather than recalled informally. A versioned definition—box, median, whisker rule, outlier treatment and aggregation unit—would let a reader interpret the graphic without deciding which paragraph to privilege. If a correction changes only the description, the manifest can record an editorial revision. If it changes the calculation, the report should receive a new analysis version and new artifact hashes. The distinction is simple once the pieces have names.

The same principle applies to the larger lineage. “2025 study” is a useful editorial label. It is not a unique key for a campaign. “Speedchecker data” identifies a collection platform. It is not a sample matrix. “Twenty origins” gives geographic coverage. It does not identify how many probes or traces represented each origin and destination. Each phrase tells the reader something; none supplies the missing join.

The manifest the report needs

The repair can be small. LACNIC could publish one immutable manifest for each report edition, readable as a table and downloadable as structured data. It should name the report identifier and edition, observation start and end, measurement campaign and dataset identifiers, collection platform, aggregate counts and counts by country or country pair, the inclusion and exclusion rule, snapshot times for any auxiliary sources, the method or code version, and durable locations or hashes for released artifacts.

Not every raw record needs to be public. Probe privacy, licence terms and security constraints can limit disclosure. Counts may be grouped, identifiers pseudonymised and sensitive fields withheld with a stated reason. The purpose is not to expose a household, a subscriber or a private topology. It is to make the boundary of a published claim inspectable.

Corrections should extend that record. An edition can say that prose changed but data did not; that a dataset was replaced; that one country pair was recomputed; or that a figure was regenerated under a corrected definition. Old manifests should remain addressable. A reader can then see whether two editions share observations and whether an apparent change comes from the network, the sample or the analysis.

This is thin coordination. LACNIC controls the edition and can attest to the evidence package it published. Speedchecker and measurement participants control their collection surfaces. Operators control their own peering, transit and routing policies. The manifest does not collapse those authorities. It gives each operator a stable reference against which to compare local telemetry.

Publication is not network reality

The broader temptation is to treat a published regional view as a current map. That is too much authority for any report. Internet paths change with policy, failure, traffic engineering, commercial agreements, maintenance and the vantage point from which a measurement begins. Even a perfectly documented observation window would describe that window, not all future behavior.

The right test is therefore not whether LACNIC can certify a region’s routing. It cannot, and it does not need to. The test is whether a reader can identify the observations behind an edition, reproduce the analytical boundary where access permits, and compare the result with local evidence without guessing which campaign was used.

LACNIC’s report already does the hard editorial work of turning many paths into a regional question. Adding the missing lineage would not weaken the story or bury it in methodology. It would let the story travel safely—from a public chart to a network engineer’s hypothesis, from one edition to the next, and from institutional publication back to the running systems that remain the final authority.

Sources

  1. LACNIC Blog: Analysis of Interconnection Paths in Latin America and the Caribbean
  2. LACNIC R&D: Analysis of Paths from Each Country to the Rest of the Region
  3. LACNIC: Spanish PDF of the regional path analysis
  4. LACNIC R&D: Earlier study of paths within each country
  5. LACNIC: English PDF of the earlier 2022 study