Summary

  • npNOG’s public record establishes continuity, programme design, participation support and institutional collaboration.
  • The same record does not yet establish durable operational outcomes; a privacy-safe outcome ledger would separate evidence, attribution and unknowns.

But continuity answers only the first question in an institutional audit: did the organisation operate? It does not answer the harder one: what changed because it operated?

Imagine a ledger with one row for every meeting. The early columns are reasonably well populated: dates, locations, tracks, speakers and, for some editions, participant counts. The later columns are much thinner: who completed the training, what was deployed, whether the change survived, and what part npNOG played rather than an employer, vendor, regulator, regional body or ordinary technology adoption.

That missing right-hand side should not be mistaken for failure. It is an attribution problem.

What the record can support

The official npNOG homepage records eleven numbered editions through 2025 and a 2020 virtual event that it says drew more than 45 participants. Its About page describes an open forum for technical and operational exchange and names npIX, ISPAN and NREN as long-term partners. Those records establish persistence, a stated constituency and an operating network of institutions. They do not, by themselves, establish network improvement. (npNOG homepage, About npNOG)

One early edition offers more concrete input measures. npIX’s account of npNOG-2 reports 328 participants overall, including 220 at the conference, and gives separate counts for routing, network-management and optical-fibre workshops. It also records seven fellowships. These are useful numbers because they move beyond a generic claim that the event was well attended. Yet they still measure access to instruction, not the use of that instruction in production. (npIX report on npNOG-2)

The 2025 record is richer in programme design. npNOG-11 published three workshop tracks covering an ISP core, enterprise networking and network automation. The curricula include practical labs, while the page states a maximum of 30 participants for each workshop. The fellowship programme offered registration, accommodation and transport support, and its results page lists seven recipients across the three tracks. (npNOG-11 workshops, fellowship, results)

This is evidence of institutional work: a programme was assembled, instructors were recruited, lab capacity was defined and some access costs were subsidised. An assessment that ignored those achievements would be as misleading as one that converted them into deployment claims.

The handoff from instruction to operation

The crucial transition happens after the closing session. A participant may understand a route-security practice but lack authority to change a production router. A fellow may promise to train colleagues but return to an employer with no time allocation. An automation lab may improve confidence without producing an adopted configuration. Conversely, a quiet conversation may resolve an operational problem with no public trace at all.

npNOG’s fellowship conditions recognise part of this transition. They ask recipients to write about their participation and to train peers. That is a sensible expectation. The inspected public pages, however, do not provide a longitudinal record of whether those steps occurred, who was reached, or whether any resulting practice persisted.

The same distinction applies to leadership. The npNOG-11 committee page identifies a Core Committee for overall management, a Programme Committee for content and a Fellowship Committee for awards. Published roles make responsibility more legible. What they do not show is a cohort history: which participants became instructors, which fellows entered committees, whether new cities or employers gained influence, and where leadership renewal stalled. (npNOG-11 committees)

What an outcome ledger would record

The answer is not a glossy impact report built from testimonials. A credible ledger would preserve several levels of evidence without collapsing them.

For each workshop, it could record enrolment, completion and a short follow-up at three, six and twelve months. A claimed deployment would need a date, an accountable network, a description that can be published without exposing security-sensitive details, and corroboration from the operator that bore the cost. A routing or RPKI trend could be included, but only with alternative explanations. A fellow who later teaches should be counted as leadership progression; an intention to teach should not.

Unknown should be a valid result. If an employer refuses disclosure, the ledger can say so. If a change cannot be separated from APNIC training, npIX work, vendor support or a wider industry trend, the record should describe contribution rather than claim causation. If an initiative produced no deployment, that finding is useful too: it may show that the barrier was budget, authority, equipment or curriculum design.

The archive itself needs the same discipline. npNOG’s homepage, Events page and npIX report contain conflicting date descriptions for an early edition. The discrepancy does not erase the event, but it shows why a long-running institution benefits from a correction log and stable records rather than silent harmonisation. (official event index)

The pattern of results should change the response. High completion with little later use would point to authority, budget or equipment constraints rather than a teaching deficit. Early deployment followed by abandonment would make maintenance, staff turnover and vendor dependence the priority. Adoption across several networks after fellows begin teaching would justify testing a peer-diffusion hypothesis, while an industry-wide shift without a dated participant trail would leave npNOG’s contribution unresolved. The ledger is useful precisely because these scenarios call for different decisions.

Informal value is real—and still deserves evidence

The strongest objection is that a network operators’ group creates trust, and trust is not captured by a spreadsheet. Engineers may call one another during an outage because they met in a workshop years earlier. A mailing-list answer may save hours without becoming a case study. Demanding disclosure can also burden participants and organisers or expose operational details.

That objection should shape the ledger, not defeat it. npNOG does not need to publish configurations, personal evaluations or private incident histories. It can report privacy-safe aggregates, attributable examples with consent, and a clear list of what remains unknown. It can sample outcomes rather than surveil every participant.

The goal is proportionality. Eleven editions are already evidence of endurance. Published curricula are evidence of serious teaching effort. Fellowships are evidence that organisers recognised access barriers. None of those claims needs to be diminished. They become more valuable when the public can see where instruction ended, where practice changed, and where the causal path is uncertain.

After a decade, npNOG’s institutional question is no longer whether it can convene another meeting. It has answered that repeatedly. The next question is whether it can preserve a fair account of what the meetings made possible—achievement, partial result and unknown alike.

Control design matters because npNOG would gain reputation from positive findings. Definitions and cohort denominators should therefore be fixed before results are known; corrections, negative findings and lost follow-ups should stay visible; and the people promoting a programme should not be the only people deciding what counts as an outcome. Otherwise incentives will favour the easiest success stories, steer teaching towards what is easiest to count, and discourage honest reports of failed deployments.

The irreversible risk is disclosure. Once a published example links a person or network to a sensitive incident, later deletion cannot restore the lost confidentiality. The safer design is to collect the minimum evidence, aggregate by default, require explicit consent for attributable cases, restrict access to raw records, and publish only the verification needed for the claim.