Summary

  • In the 1994 RIPE NCC Contributors Committee minutes, Karrenberg argued that staffing should be driven by growth in demand for NCC services rather than by Internet traffic growth; the recorded conclusion was to endorse personnel expansion and begin recruitment.
  • His 1993 first-year report had already described the NCC in service terms—number assignments, database maintenance, information access, meeting support and technical-project hosting—providing a foundation for treating operational demand as something different from the growth of the Internet as a whole.

The argument that changed the denominator

The revealing moment in the 1994 RIPE NCC meeting lasts only a few lines.

Daniel Karrenberg, the NCC manager, had presented what the minutes call the organisation’s “workload situation.” Local registries were growing at a rate the minutes described as nearly exponential. The centre had 3.5 full-time-equivalent staff, while Karrenberg’s assessment put the existing workload at five FTE. Activities associated with a routing registry would require another FTE. His proposed intermediate requirement for the last quarter of 1994 was therefore six FTE, with at least one additional engineer during 1995 if growth continued. He also described an operational asymmetry: every provider created a base level of interaction, larger providers added variable work, and new providers often required more help because they had to be brought up to speed. At the current staffing level, the training he considered necessary for newcomers could not realistically be organised. These details are recorded in the 1994 meeting minutes.

The ensuing exchange sharpened the question. Peter Dawe thought the staffing plan was insufficiently ambitious. Glenn Kowack observed that RIPE NCC personnel could not increase at the same exponential rate as Internet traffic. Karrenberg answered by rejecting traffic as the operative denominator: according to the minutes, the problem was not growth in traffic but growth in demand for NCC services. Staff therefore had to increase if service quality was to be maintained.

The distinction matters because the two quantities are not the same thing.

Internet traffic can rise because existing networks carry more packets. The number of customers attached to networks can rise. The number of autonomous networks can rise. Capacity can increase. Applications can become more data-intensive. None of those developments maps mechanically to the amount of work arriving at a coordination centre.

A service organisation instead experiences demand through its own interfaces. Requests arrive. Database records need processing or maintenance. New local registries establish relationships with the centre. People need information, documentation or assistance. Meetings need support. New activities create new queues or specialised tasks. A rapidly growing Internet can make all of these things more likely, but it does not determine them one-for-one.

Karrenberg’s intervention therefore changed the staffing equation. The question was no longer, “How can an organisation grow as quickly as the Internet?” It became, “What work is actually reaching this organisation, what effort does that work consume, and what capacity is required to provide the agreed services at an acceptable quality?”

That is a narrower proposition, but a much more manageable one.

It is also the useful centre of gravity for understanding Karrenberg’s role in the RIPE NCC’s early years. The organisation’s own speaker biography describes him as a founder, an early European Internet pioneer, RIPE NCC CEO from 1992 to 2000 and Chief Scientist from 2000 to 2013. Those retrospective labels describe a long career. The 1993 and 1994 operating documents reveal something more concrete: how a person in charge of a young coordination centre described its work while the work was becoming too large for its initial staffing model.

The important unit was not “the Internet.” It was a service.

Before capacity planning, there had to be a service organisation

The RIPE NCC did not begin by inventing every activity it performed.

Karrenberg’s 1993 report, The First Year of the RIPE NCC, says that basic services were available from the organisation’s first day because organisations that had previously performed the work voluntarily provided substantial support. That sentence is easy to pass over because institutional histories naturally emphasize the moment at which an organisation formally begins. Operationally, however, continuity mattered more than novelty.

The work already existed.

The transition was from work carried through voluntary arrangements toward work placed inside a dedicated centre with named staff, an activity plan, a budget and recurring reporting.

The 2017 RIPE NCC retrospective describes the motivation in similar institutional terms: members of the early RIPE community concluded that the Internet would not remain a small experiment and that a dedicated structure was needed for coordination activities in Europe. The retrospective says the NCC began formal operations on 1 April 1992 at NIKHEF in Amsterdam with Karrenberg as manager. It is the RIPE NCC’s own later account and should be read as such, but it is consistent with the contemporary 1993 report.

That report gives the initial operating arrangement more texture. Karrenberg says the NCC began in offices rented from NIKHEF in April 1992 with him as manager. Anne Lord joined in May as network administrator, along with Marten Terpstra as network engineer. Financial and legal administration was provided by the RARE secretariat.

This arrangement is significant because it already separated kinds of responsibility.

There was technical and operational work inside the NCC. There was financial and legal administration supplied through another organisation. There was a broader RIPE technical community. And there were organisations that had previously carried some of the relevant work voluntarily.

Karrenberg stood at an important junction among those parts, but the documentary record does not support reducing them to one person.

That separation becomes even clearer later. In 1994, the planned organisational model said that the NCC manager would produce a work plan using RIPE technical expertise as input; a new organisational umbrella would determine the budget necessary to provide the service; and the Contributors Committee would give required advice on the minimum set of services and on the tariff structure. The same 1994 minutes record the committee accepting that proposed organisational scheme.

The manager proposed and operated. The technical community supplied expertise. The legal and operating umbrella had its responsibilities. Contributors had a defined advisory role. Later governance structures would change, but even in this transitional form the evidence resists a heroic-founder model in which institutional authority and operational responsibility are imagined as identical.

That distinction is central to Karrenberg’s profile because the most interesting part of his work in these records is not that he possessed every form of authority. It is that he had to convert one kind of evidence—operational demand—into a case that other participants and organisational actors could evaluate.

A first-year report built out of service counters

The 1993 report is short. Its compactness makes its choice of measurements especially revealing.

Karrenberg reported that the NCC had established the European Internet Registry, assigning Internet numbers to European organisations through a function distributed among local registries. He wrote that more than 10,000 IP network numbers had been assigned during the first operating year. He said the RIPE network-management database was maintained and extended, with more than 500 database updates processed by the NCC on an average working day. The centre maintained information for service providers and network operators in the RIPE document store and made it available through interactive access and contemporary Internet information-discovery methods; the report counted more than 200,000 queries during the year. It also described clerical support for RIPE working groups and organisation of RIPE meetings, as well as the NCC providing a home for separately funded technical-development projects identified jointly by RIPE and RARE. These claims all come from Karrenberg’s first-year report.

Those numbers are useful, but only if they are allowed to remain what they were: service counts reported by the manager of the service organisation.

“More than 10,000 IP network numbers” describes assignments, not traffic.

“More than 500 database updates” on an average working day describes processing volume, not the number of Internet users, not global network reliability and not the quality of every database record.

“More than 200,000 queries” describes use of information services, not the number of distinct people served and not proof that every query produced a satisfactory outcome.

The report does not independently establish the accuracy of every underlying transaction. It does not measure user satisfaction systematically. It does not demonstrate broad representation merely because a service was used. It does not provide a controlled performance study. It was written by Karrenberg as NCC manager to summarise the NCC’s own first year.

Yet precisely because those limitations are clear, the counters remain useful.

They show what the operator thought was worth counting.

If the institution’s purpose had been described only at the altitude of “supporting the Internet,” the phrase would have been too elastic for serious capacity planning. The first-year report instead decomposed that ambition into activities: assignments, updates, queries, documents, meetings, technical projects. Once work has interfaces and counters, staffing questions can be attached to something more concrete than the size of the surrounding network.

That does not mean the counters were perfect proxies for labour.

A database update may be trivial or complicated. Two assignment requests can demand very different levels of review. A query can be answered automatically at negligible marginal labour, while helping a new provider understand procedures can consume sustained staff time. Organising a meeting is difficult to compare numerically with processing a registry request. Software and documentation can change the amount of human effort associated with an apparently similar transaction.

In other words, demand is not identical to transaction count either.

But it is closer to the operational problem.

The 1994 minutes show Karrenberg adding qualitative structure around that problem. There was “base workload” in interactions with providers, the minutes say, while variable workload depended on provider size. New providers imposed additional work because they often needed education. A routing-registry activity added another category of labour. That is already more sophisticated than multiplying Internet traffic by a staffing ratio.

It treats workload as a composition.

What service demand can reveal that traffic cannot

Consider the difference between a network transport problem and a coordination-service problem.

If a backbone operator observes rapidly rising traffic, capacity planning may sensibly begin with bits per second, interface utilisation, link headroom, traffic matrices and failure scenarios. Those metrics are close to the mechanism by which the operator delivers its service. More traffic can directly consume more transport capacity.

The RIPE NCC’s work described in the 1993 report had different mechanisms.

A request for Internet number resources could require registry interaction. A database update could require systems and operational processing. A new local registry could require guidance. A technical activity could need engineering. A working group could require clerical support. A meeting could require organisation. A new registry service could introduce both implementation work and continuing operational obligations.

Packets crossing the Internet did not pass through Karrenberg’s staff as a corresponding stream of manual tasks.

That sounds obvious now. In 1994 it was important enough to say aloud because exponential Internet growth was a dominant fact of the environment. When a system around an institution expands rapidly, its most dramatic growth curve can become an attractive but misleading planning metric.

The 1994 exchange shows resistance to that shortcut.

Kowack’s observation, as recorded, was reasonable on its own terms: personnel could not simply track exponential traffic growth. Karrenberg’s response did not dispute that constraint. Instead, it separated the external growth indicator from the institution’s direct workload.

This makes the recorded conclusion more intelligible. The committee did not endorse a rule that NCC headcount should track Internet traffic. It endorsed the personnel-expansion plan that had emerged from the workload presentation and said recruitment should begin.

There is a broader managerial pattern here, but it should not be romanticised. Operators everywhere try to identify the variable that constrains their service. Sometimes they choose well; sometimes they choose badly. Contemporary evidence here does not prove that every staffing estimate Karrenberg made was exact. His statement that existing work required five FTE, for example, appears in the minutes as his presentation, not as an independently audited time-and-motion study.

The value of the record lies in the category distinction.

He identified a falsifiable operational question: how much demand was being placed on the NCC’s services, and what staffing did the centre need to meet that demand without degrading quality?

That formulation creates room for correction. Work can be measured differently. Processes can be redesigned. Automation can reduce labour. Service definitions can change. Training can alter future support demand. Additional activities can be accepted or deferred. Staffing assumptions can be revised.

“Internet traffic is growing” supplies far fewer such levers.

Service quality is the missing term

Karrenberg’s recorded argument did not end with demand. It connected demand to quality.

That matters because staffing need is not determined by workload alone. It is determined by workload relative to the level of service the organisation intends to sustain.

An overloaded operation has several possible responses. It can add capacity. It can reduce scope. It can automate. It can tolerate longer queues. It can lower the amount of assistance given to each requester. It can alter eligibility or procedures. It can prioritise some work and defer other work.

The 1994 minutes record Karrenberg advocating one response: increasing staff so that quality of service could be maintained.

The document does not provide a modern service-level dashboard. It does not define “quality” through a single response-time target, error threshold or satisfaction score. That absence should constrain interpretation. There is no basis in this source set for claiming that Karrenberg had already built a comprehensive quantitative quality-management regime.

But the minutes contain a useful operational example.

New providers, Karrenberg said, often required education, and a course for newcomers was necessary; with the existing staff, however, organising such a course would be effectively impossible. Here quality is not merely throughput. It includes an ability to assist participants who do not yet know how to use the service relationship efficiently.

That creates a feedback mechanism.

If new providers need substantial one-to-one help, they consume staff capacity. If training can make new providers more capable, a training programme may eventually reduce some support burden or improve the quality of interactions. But developing and delivering training itself requires capacity. An understaffed organisation may therefore be too busy handling demand to invest in mechanisms that make future demand easier to handle.

The meeting minutes do not demonstrate that this feedback loop played out exactly that way. They do, however, show Karrenberg identifying the immediate version of the constraint: newcomer education was desirable and current staffing made it impractical.

This is a more interesting leadership problem than headcount in isolation.

Capacity has to cover both today’s queue and tomorrow’s improvements.

A service institution that staffs only for visible transactions can starve documentation, training, software development and process improvement. The immediate queue gets processed until accumulated complexity makes each future request harder. Conversely, an organisation that expands every supporting function without relating it to demonstrated service needs can build capacity faster than its obligations justify.

The operator’s problem is to keep those two errors apart.

The 1994 record captures Karrenberg arguing from the first risk: service demand was already above the capacity represented by 3.5 FTE, in his assessment, and planned new work increased the requirement further.

The routing registry demonstrates why scope matters

The proposed routing registry is especially helpful because it shows how workload can increase without any need to invoke aggregate traffic.

In the 1993 report, Karrenberg identified establishment of a routing registry as a major new activity for the following year. He described the intended function in practical terms: Internet service providers would be able to register their routing policies and retrieve information about other operators’ policies, with tools intended to help routing stability.

By the 1994 staffing discussion, the activity appeared in workload planning as requiring another FTE.

Whether that particular estimate proved accurate is not established by the source set. The conceptual point is firmer: a service can create workload because the institution chooses, through its planning and governance processes, to undertake it.

This distinguishes three phenomena that are often collapsed.

The Internet can grow.

Demand for existing institutional services can grow.

The institution’s own service scope can grow.

All three can increase workload, but through different paths.

Traffic growth is an environmental fact. Existing-service demand is a usage fact. Scope expansion is partly a decision.

A manager who does not distinguish them can make accountability difficult. If every added activity is described merely as an inevitable consequence of Internet growth, discretionary choices disappear from view. If every workload increase is treated as a management decision, genuine external demand is understated.

The 1993 and 1994 documents make a cleaner distinction possible. Karrenberg reported continuing services, demand associated with local registries and providers, and an additional staffing need tied specifically to a planned routing-registry activity.

That is operational accounting in the broader sense: not just money, but causation.

From voluntary work to an institution that could say what it did

One of the most consequential transitions in the first-year report is barely visible because it had already happened before the document was written.

The NCC inherited work from organisations performing it voluntarily. Its creation did not make coordination necessary; it made parts of that coordination legible as the responsibility of a continuing service organisation.

This changed what could be asked.

Who is responsible for processing the work?

How much arrives?

What functions are core?

What needs dedicated engineering?

What is separately funded?

How many staff are available?

What new activities are being contemplated?

What happens if demand rises faster than capacity?

These questions become easier to pose when a named organisation accepts operational responsibility.

The 1993 report illustrates that shift through its mixture of service descriptions, statistics, staffing and expenditures. Karrenberg reported that the NCC had met the main goals of its activity plan and that a technical review was underway within RIPE. He also listed expenditures against budget. The article here need not turn into a history of fees or reserves to see why this matters: a planned service organisation has to connect promised activity to resources.

The report’s operating section also prevents the story from becoming one of purely personal accomplishment. Lord and Terpstra were named members of the initial staff. RARE supplied financial and legal administration. RIPE and RARE jointly identified technical-development projects. Previous organisations enabled basic services to start immediately.

Karrenberg managed an assemblage before it became a more mature institution.

That is a different kind of leadership from unilateral control. It requires describing dependencies well enough that the work can persist beyond individual relationships.

The 1994 committee did not simply ratify a manager’s authority

A person-centred article can distort institutions by treating every documented proposal from its subject as a decision. The 1994 minutes are useful precisely because they preserve the distinction between presentation, discussion and conclusion.

Karrenberg presented his workload analysis and staffing requirements.

Dawe and Kowack responded.

Karrenberg made the service-demand distinction.

The minutes then separately record the conclusion: the personnel-expansion plans were endorsed and recruitment should begin.

Immediately afterward, the meeting moved into the proposed organisational scheme. Kees Neggers presented that proposal, not Karrenberg. Under it, the NCC manager would prepare a work plan using RIPE technical expertise. The new operating organisation would determine the budget required to provide the service. The Contributors Committee would provide prescribed advice on the minimum service set and tariff structure. After discussion, the minutes state that the committee accepted the organisational scheme.

The differences among those verbs are important.

Presenting is not endorsing.

Operating is not deciding every institutional question.

Providing technical input is not identical to holding legal authority.

Funding a service is not the same as performing it.

Participating in a committee is not the same as individually controlling the outcome.

The historical record becomes more useful, not less, when these distinctions are retained.

The later first RIPE NCC Executive Board minutes from 1997 show a more formalised version of institutional separation. The board unanimously appointed Daniel Friedrich Karrenberg to the function of RIPE NCC General Manager under the statutes. The same minutes established interim financial authorisations, required two-person sign-off for payments under the procedures described there, approved agreements presented by management and dealt with the transfer of assets and liabilities from TERENA to the association.

Those records demonstrate substantial managerial responsibility. They also demonstrate a board taking resolutions, setting authorisations and approving documents.

The distinction is not semantic housekeeping. It tells us what kind of evidence is needed to attribute an institutional action.

A Karrenberg recommendation can be attributed to Karrenberg.

A committee conclusion belongs to the committee record.

A board resolution belongs to the board.

The activities of RIPE, the NCC, its organisational host and its contributors should not be folded into whichever individual is most visible in the archive.

A career that repeatedly moved between building and handing off

The RIPE NCC’s current official biography for Karrenberg describes his later trajectory in broad terms. It says he served as CEO from 1992 to 2000 and Chief Scientist from 2000 to 2013, and credits him with leadership in activities including the Routing Information Service, RIPE Atlas, DNSMON, RIPEstat and NSD with NLnet Labs. It says he now supports the CEO and RIPE Chair Team on a part-time basis.

Because this is an organisational biography, its claims of leadership and pioneering status are the RIPE NCC’s characterisation of its own long-serving figure. They are useful for chronology, but they should not substitute for the operational records where those are available.

The 2022 recognition notice similarly tells the history through institutional commemoration. It says Karrenberg helped start the RIPE community with Rob Blokzijl in 1989, became the RIPE NCC’s first managing director after helping establish it in 1992, passed leadership to Axel Pawlik around the turn of the century, and then led research-and-development efforts and served as Chief Scientist.

Again, the important word for this article is not “pioneer.” It is succession.

A coordination institution proves something about itself when responsibilities that were closely associated with an early operator can be transferred.

The 2013 senior-management notice provides a later example. It announced that Kaveh Ranjbar would become Head of the Science Division and that Karrenberg would step down from that role while remaining as an adviser to the managing director. The organisation said Karrenberg had accepted operational charge of the division on the condition that it would be temporary, although the period became longer than expected. Ranjbar was to build on that work.

That notice is not evidence about how efficiently every handoff occurred. It is evidence that the organisation publicly described a transfer of operational responsibility rather than an indefinite concentration of it.

For an institution whose early story depended heavily on a very small staff, this matters.

Capacity is not only the number of people available. It is also the distribution of knowledge and responsibility among them.

An organisation can have sufficient headcount and still be fragile if only one person understands why a process exists, how exceptions are handled, which external relationships matter, or what historical constraints shaped a service. Conversely, documentation, delegation and explicit role changes can increase practical institutional capacity even without increasing staff.

The early workload record and later succession notices therefore belong in the same profile. Karrenberg’s observable contribution was not simply accumulation of functions. At several stages, the record also contains handoffs.

Thirty years later, participation still did not equal decision authority

A particularly useful late-career document is RIPE-849, the 2025 RIPE Nominating Committee final report, edited by Karrenberg.

It concerns a very different function from the registry workload of 1994, but it provides a clean example of role separation.

The report says the Nominating Committee selected the RIPE Chair and Vice Chair through a process involving ten randomly selected voting members, a chair, advisers and liaisons. Only the voting members voted on candidate selection. Karrenberg served in a non-voting role as previous chair and adviser. The report says he helped the chair draft the committee timeline and rules, established supporting systems, and offered to serve as scribe for minutes and editor of the final report so that additional RIPE NCC clerical support would not be needed.

That is participation. It is real work. It includes infrastructure, drafting, record keeping and institutional memory.

It is not the vote.

The example helps illuminate the earlier history because coordination institutions repeatedly depend on work performed by people who do not, merely by performing that work, acquire every relevant decision right.

Someone maintains the record.

Someone operates the service.

Someone provides evidence.

Someone advises.

Someone votes.

Someone bears legal responsibility.

Sometimes one person occupies several of those roles, but the roles remain analytically separable.

Karrenberg’s 1994 staffing argument fits this pattern. He had operational responsibility for presenting the NCC’s workload and recommending the capacity he thought necessary. The committee discussed and endorsed the expansion. Other organisational actors had their own responsibilities.

That is a stronger account of leadership than attributing everything to a founder, because it shows the actual work by which an institution converts technical needs into authorised action.

The limits of the first-year numbers

The temptation with old operating reports is to treat any surviving metric as more complete than it was.

Karrenberg’s 1993 figures deserve more care.

The number-assignment figure tells us that a substantial quantity of assignment activity was reported during the year. It does not by itself tell us how difficult the requests were, how evenly they were distributed over time, how much was handled locally rather than centrally, or how recipients evaluated the service.

The database-update figure tells us about reported processing volume. It does not establish the semantic quality of every update or how much operator intervention each one required.

The query figure demonstrates reported use of information systems at significant scale for the young service. It does not identify unique users. A small population could generate many queries. Automated access can magnify counts. Query volume need not correlate linearly with staff workload.

The meeting and clerical functions are even harder to compress into transaction counts. Their labour can involve preparation, scheduling, communication and follow-up whose cost is not obvious from a single headline metric.

The report itself does not claim otherwise.

Its purpose was to summarise a year, not provide the kind of workload model that would later be expected from a mature service organisation.

That incompleteness makes the 1994 narrative evidence more valuable. Karrenberg did not simply say, “Transactions increased, therefore staff must increase.” The recorded presentation added relationships between provider count, provider size, newcomer support and new service activities.

That is closer to a causal model.

It remains his model, reported in meeting minutes. The committee’s endorsement does not transform every assumption within it into an independently proven fact. But the exchange shows that staffing was being argued through workload mechanisms rather than through a generic appeal to Internet importance.

Why the traffic distinction remains conceptually durable

The Internet makes aggregate numbers unusually seductive.

Traffic volumes are tangible. Prefix counts are tangible. User counts are tangible. Membership totals are tangible. Growth charts look authoritative because they move.

But an institution can drown in a workload that barely registers in the most visible aggregate metric, or remain lightly loaded while the surrounding system expands rapidly.

The correct planning variable depends on the service.

A policy process may be constrained by review complexity rather than packet volume.

A registry may be constrained by the quantity and difficulty of requests.

A measurement platform may be constrained by hardware, software, storage, network capacity, analysis or support in different proportions.

A training function may be constrained by instructor time and participant demand.

A coordination meeting may be constrained by agenda complexity and human attention.

The enduring value of Karrenberg’s 1994 statement is not that “service demand” is always the right single metric. It is that the institution must identify the variable that actually consumes its capacity.

Even service demand needs decomposition. One type of request may become almost fully automated while another remains labour-intensive. New users may create disproportionately high support loads. A new service may require engineering before it generates visible external demand. Reliability work may consume substantial effort precisely because users do not experience incidents.

No one metric solves this.

The discipline is to avoid substituting the growth of the surrounding environment for evidence about the organisation’s own work.

Leadership as translation

Profiles of Internet pioneers often gravitate toward invention, firsts and institutional birth.

The documents here support some of that vocabulary. The RIPE NCC’s biography calls Karrenberg a pioneer and founder. Its 2017 retrospective places him at the opening of formal operations. Its 2022 recognition notice commemorates his long tenure.

But the less ceremonial records show a more instructive activity.

Karrenberg translated.

He translated a collection of inherited coordination functions into an operating report.

He translated activity into service categories.

He translated parts of those services into counts.

He translated growing relationships with registries and providers into workload.

He translated a planned routing function into an incremental staffing requirement.

He translated insufficient capacity into a service-quality risk.

Then other people and organisational bodies could argue with the result.

That last step is essential.

A useful operational model is not merely something a leader announces. It is something others can inspect, question and ultimately accept, reject or modify through whatever authority structure applies.

The 1994 minutes preserve exactly that sequence. Karrenberg gave his estimate. Dawe challenged it from one direction, saying the staffing ambition was too low. Kowack introduced another concern, rejecting any simple attempt to follow traffic growth with personnel. Karrenberg refined the denominator. The meeting reached a recorded conclusion.

This is how technical institutions often become governable: not by eliminating expert judgment, but by making expert judgment specific enough to be discussed.

Maintaining a service is different from owning the system it serves

There is a final distinction worth making because Internet institutions are unusually prone to being described in sovereign language.

The RIPE NCC’s services mattered because networks depended on coordination. Karrenberg’s first-year report itself called those services essential to the Internet, and the organisation’s retrospectives understandably emphasise their importance.

Operational importance, however, is not the same as ownership of the Internet.

The 1993 report presents a distributed registry function involving local registries. It describes support for RIPE working groups rather than rule by the NCC over those groups. It names RIPE and RARE as jointly identifying technical projects. It acknowledges prior voluntary work. The 1994 organisational model divided planning, technical input, organisational responsibility and contributor advice. The 1997 board minutes show formal governance outside the general manager himself.

The Internet that generated demand for these services was composed of independently operated networks and organisations. The NCC coordinated specific functions within that environment.

That boundary is important precisely because infrastructure coordination can be powerful.

When a service becomes widely relied upon, operational decisions matter. The people who design processes, maintain records and allocate scarce organisational attention can influence what is easy, slow, visible or difficult. That influence should be described concretely rather than inflated into a claim of general authority.

The early documents make concrete description possible.

Karrenberg was responsible for operating and explaining services.

He was not, by virtue of that responsibility, the Internet’s executive.

The institutional lesson in six FTE

The number that stands out in the 1994 minutes is six.

Karrenberg said 3.5 FTE were available. He assessed the current work at five. A routing-registry activity would add another. Six FTE were therefore needed in the last quarter, with another engineer anticipated during 1995 and further personnel possible if registry growth continued.

There is no reason to fetishise the exact number three decades later. Staffing estimates age quickly, and the archival record provided here does not audit the assumptions behind each increment.

What matters is what had to happen before “six” could be proposed.

The organisation had to define services.

It had to observe demand.

It had to distinguish recurring workload from variable workload.

It had to account for new users who imposed additional support requirements.

It had to identify new activities separately.

It had to connect inadequate capacity to a consequence: declining service quality.

And it had to put the recommendation before people who could discuss and endorse it.

This is infrastructure management at a scale small enough to see its components.

The later RIPE NCC became much larger and developed many more services. Karrenberg’s own roles changed: CEO, Chief Scientist, science-division leader, adviser and participant in community processes, according to the organisation’s biography, 2013 management notice and 2022 recognition record. By 2025, RIPE-849 shows him supporting another governance process in an explicitly non-voting capacity.

The striking continuity is not personal authority. It is attention to the machinery that lets collective work happen: service definitions, systems, records, roles and handoffs.

The 1994 traffic exchange is valuable because it exposes the reasoning underneath that machinery.

Internet traffic could grow exponentially without telling Karrenberg how many people the NCC needed. The organisation needed to measure the work arriving at its own doors.

Once he made that distinction, staffing ceased to be an attempt to keep pace with the Internet as an abstraction. It became a question about an institution’s promises: what services it had agreed to provide, what demand those services were attracting, and how much operational capacity was necessary to provide them without allowing quality to deteriorate.

That is a modest definition of leadership compared with the mythology often attached to Internet founders.

It is also the one these documents support.

Sources