Summary

  • RIPE NCC’s June rebuild moves LatencyMON to Vue 3 and Quasar and extends the view across ping, traceroute, DNS, TLS and HTTP measurements.
  • Its documentation distinguishes absolute and relative axes, min-only and min/middle/max views, country-size and target-distance grouping, aggregated and per-sample retrieval, and periodic polling rather than streaming.
  • Those choices make a visualisation useful. They do not establish an outage, a cause, a route outcome, an affected operator or a conclusion about the Internet beyond the chosen measurement view.

A better window is not a completed record

The rebuilt LatencyMON solves a real reading problem. One line for every RIPE Atlas probe becomes unreadable quickly. The tool groups probes, draws a band, and lets a reader compare latency over time. The June announcement says the new version can begin with a fuller aggregated view, add DNS, TLS and HTTP beside ping and traceroute, group by countries with the most probes or by distance from the target, and split a group back into individual probes.

Those are choices about what a reader sees. The current documentation makes the choices more precise. A ping chart uses an average as its middle value; traceroute, DNS, TLS and HTTP use a median. The axis may be milliseconds or a relative percentage above the lowest observed latency. The display may retain the full minimum–middle–maximum band or reduce it to the minimum. A group may represent country, target or a reader-selected set of probes.

None of those facts makes the picture unreliable. They explain why two accurate views of related measurements can look different. A narrow minimum line answers a different question from a full spread. Relative mode makes movement visible across unlike baselines; it does not preserve the same numerical meaning as milliseconds. A country group is not a claim about every network in that country.

The data path is part of the meaning

LatencyMON normally reads an aggregated result path called routequake. Outside that range, it can retrieve underlying non-aggregated per-sample results, with a current cap of 10,000 samples for performance. Recent windows can poll for new results at the measurement’s sampling interval; the documentation explicitly says this is not a streaming feed.

That is enough to identify another boundary. An aggregate band, a capped raw slice and a periodically refreshed view are legitimate tool states, but they are not interchangeable evidence objects. The published documentation does not say that any path loses data, that a visible value is wrong, or that a particular historical conclusion cannot be reproduced. It says what the tool can load and how it presents it.

A small interpretation receipt

A shared link or embed already preserves a great deal of view state. The documentation says the URL fragment carries the measurements, time window and settings, and the tool links back to underlying measurement IDs. That is a strong starting point. When a chart is cited in an operational discussion, the surrounding record should make its evidential scope legible too.

A compact interpretation receipt could preserve the measurement IDs; the shared link or a state digest; the selected time range and timezone; absolute or relative mode; min-only or all-values mode; grouping basis and visible probes; aggregate or per-sample retrieval; retrieval time; the 10,000-sample cap if it applied; and a link back to the measurement. It would state whether the middle line means average or median.

This is a proposed reading aid, not a requirement RIPE NCC has announced. It should not expose probe owners, targets, credentials or private observations. Nor should it pretend that an interface record proves a service incident. Its purpose is narrower: another reader can see what the chart was asked to represent before treating it as evidence.

What the sources do and do not establish

RIPE NCC’s sources establish a rebuilt visualisation and its documented display and retrieval behaviour. They do not identify a real measurement, target, outage, network fault, customer impact, service-level breach or operator decision. They do not show that a particular shared view lacks provenance or that RIPE NCC must implement this receipt.

Sources