Summary

  • An IETF charter revised on 1 October proposes merging the Benchmarking Methodology and IP Performance Measurement groups into a new Benchmarking & Performance Measurements group. BPM remains proposed, in internal charter review; it is not yet an approved working group.
  • The latest text separates controlled laboratory benchmarks from measurements on live, variable networks. It preserves an older limit: the group may define how results are obtained and reported, but its benchmark recommendations will not set acceptance criteria or performance requirements.

A network test can produce a precise number and still leave the decisive question unanswered. Was it generated by repeatable stimulus in an isolated lab, or measured while customers and routing policies changed on a production network? The proposed BPM charter now says those are different conditions, not interchangeable sources of a performance verdict. The distinction matters because the IETF is considering one group to inherit both the controlled benchmarking work of BMWG and the operational measurement work of IPPM.

The version dated 1 October adds explicit language about the two environments. A laboratory assessment isolates and controls its stimulus to characterize devices, systems or services. Production measurement observes a service or application under live, variable operating conditions. The previous charter version already discussed labs and production, so this is a clarification of the divide, not the invention of either practice. BPM's proposed scope spans data, control and management planes, including physical, virtual and containerized infrastructure.

The charter also preserves a limit that is easy to lose in a product claim. Benchmark recommendations can specify relevant performance characteristics, a test environment and unambiguous reporting. They do not define a passing score or a minimum performance requirement. That exclusion was already in the 29 September version. A purchaser can use a standard method to compare two results and then set its own acceptance threshold; the proposed IETF group would not choose the threshold for that purchaser. Reproducibility is evidence about the test, not authority over the procurement or operational decision.

The institutional change is still conditional. The proposed text calls BPM the successor to BMWG and IPPM and would give it maintenance of their RFCs and performance-metric guidance under BCP 170. Its Datatracker page still marks the group Proposed and the charter in internal IESG/IAB review, with an unresolved BLOCK displayed at inspection. A table of proposed submission dates, including work on telemetry and source-address-validation benchmarking, is a plan, not a set of published standards. The earlier groups should not be described as already closed.

The review record exposes why a merger of methods can also become a dispute about mandate. In a September IETF list exchange, Gunter Van de Velde questioned language that might have given BPM an IETF-wide coordination role beyond the predecessor groups' collaborative functions. The 1 October charter speaks instead of collaboration with groups responsible for adjacent data-plane work. That is an attributed concern and a textual revision, not a finding that BPM seized authority.

Its technical boundary still needs to be read alongside the distinction between a measurement protocol, an encapsulation owned elsewhere, and a decision made by an operator.

One more revision illustrates why the exact version matters. A paragraph in the 29 September text generally expected implementation evidence for Standards Track work and specified how exceptions would be justified. That paragraph is absent from the 1 October version. The charter alone does not explain the deletion; it would be unsound to infer that the IETF has broadly abandoned implementation discipline. What can be said now is narrower: the proposed home for measurement has redrawn its lab/production language and left the pass/fail decision outside its benchmark recommendations, while its own approval remains open.

Sources