Asia-Pacific Scale Before Universal Broadband

Summary

• APNIC's early constituency was shaped by reachable research networks, national registries and service providers before the region's later mass Internet use existed.
• The 1994 pilot figure of 27 members in 12 economies, the April 2000 figure of 396 members in 34 of 62 described economies, and today's 56-economy regional statement are three different institutional denominators, not one clean time series.
• Population, individuals using the Internet, and APNIC institutional entities answer different questions; none can be silently substituted for the others.
• The regional registry boundary is justified mainly by unique-ledger efficiency, not population parity, but demographic and broadband mismatch still tests meeting access, service latency, fees, language reach and practical participation.

The mismatch begins before mass connectivity

APNIC's early scale problem begins with a simple asymmetry. The region that would later contain huge numbers of Internet users was not, in the early 1990s, a region of universal broadband or mass commercial access. It was a region where the visible Internet geography ran through research networks, university links, national Internet registries, technical coordinators and early service providers.

That starting point matters because institutions often inherit weight from the networks that are connected first. The first groups to need address allocation, maintain international circuits, attend regional meetings and understand global registry rules are not necessarily the groups that later carry most users, most population or most economic dependence. Early connectedness creates administrative presence. Administrative presence can become membership. Membership can become voting weight, meeting attendance, policy fluency and staff familiarity.

Those channels can persist long after the public Internet has expanded far beyond the original circle.

This article does not argue that APNIC should have allocated number resources by population. It should not have. A regional Internet registry coordinates a unique ledger for network operators. It does not distribute addresses as a demographic entitlement. A country with many residents but few requesting networks does not automatically need more registry votes or addresses than a smaller economy with many operators, networks or resource holders. Population is the wrong operational rule for address allocation.

But population and Internet-use data are still important context. They test whether institutional participation keeps pace with the wider region affected by the registry's work. If the early connected map becomes durable institutional weight, then later public scale raises governance questions: who can attend, who can afford participation, who can use local-language support, who receives timely service, who appeals decisions, and whose operators become visible in the membership record?

The evidence here is deliberately bounded. APNIC's pilot history reports 27 members in 12 economies in 1994. APNIC's April 2000 status report gives 396 members in 34 of 62 then-described economies. APNIC currently describes a 56-economy region. World Bank and ITU data provide independent Internet-use and broadband context, but the fixed record used here does not supply a complete economy-year dataset for all members, voters, attendees, resource holders, requests or appeals from 1993 to 2010. The right conclusion is therefore a denominator audit, not a causal verdict.

The question is not whether APNIC's early map was illegitimate because it was small. The question is how far a small connected founding map shaped institutional weight before the region's later public Internet scale existed.

Three denominators, three different questions

The first denominator is population. It counts residents. Population is useful for understanding the human scale of a region, the potential social importance of Internet access, and the mismatch between early technical participation and later public dependence. It does not identify address demand. It does not identify network operators. It does not identify APNIC members or voters. It is a public scale denominator, not a registry entitlement denominator.

The second denominator is individuals using the Internet. World Bank data sourced from ITU give country-year percentages of people using the Internet. ITU's 2010 global facts and figures provide end-period adoption context for comparing the early 1990s with the mass-connectivity period. This denominator is closer to Internet dependence than population alone, but it still does not identify who needs number resources directly. Users consume connectivity through providers, institutions, mobile networks, enterprise networks and public access points. They are affected by registry quality, but they are not the ordinary applicant class.

The third denominator is APNIC institutional participation. This includes members, resource holders, voters, meeting attendees, Executive Council candidates, national registries, service providers and operators visible in APNIC's own records. It is the denominator most relevant to internal governance. It is also the one most likely to preserve early connectedness, because membership and resource holdings arise from the networks that were active and organised enough to appear.

The three denominators answer different questions. Population asks: how large is the public affected in the long run? Internet-use percentage asks: how widely has network dependence spread? APNIC participation asks: who is inside the institution's service and governance machinery? A number from one denominator cannot answer a question about another. A member count cannot prove user representation. A user percentage cannot prove address demand. Population cannot prove voting entitlement. Economy coverage cannot prove operator coverage.

This separation is essential because APNIC's early and later records use institutional denominators. Twenty-seven members in 12 economies is not a population statistic. Three hundred ninety-six members in 34 of 62 described economies is not a user statistic. A current 56-economy boundary statement is not a continuous 1993-to-present dataset. Each figure is meaningful, but only inside its own frame.

The governance problem begins when institutional figures are allowed to borrow emotional force from population or user scale without being measured against them. Saying that a registry covers a vast region can imply a scale of public responsibility. Saying that it has members across many economies can imply representation. Those are related but not identical claims. The hard work is comparing them without converting one into the other.

The first map was a map of reachable institutions

The earliest APNIC constituency was not a demographic map. It was a map of reachable Internet institutions. APNIC-hosted history on research and education networks identifies PACCOM and early university or research-network links in places such as Australia, Hong Kong, Japan, Korea and New Zealand. Those examples help explain why the early Asia-Pacific Internet was first visible through research and technical institutions rather than through mass-market broadband.

That first map was practical. The networks with international circuits, technical staff and contact with global coordinators were the networks able to bring registry problems into a regional conversation. They were also the networks most likely to understand why a unique address ledger mattered. If a regional registry had to be built in the early 1990s, it would almost certainly be built from those institutions rather than from a later mass-user base that did not yet exist.

The pilot history's figure of 27 members in 12 economies in 1994 should be read against that setting. It is evidence of a real multi-economy institutional base. It is not evidence that the region's population or future Internet users were represented. It is not even a full network-operator census. The denominator is the pilot membership and economy presence recorded by APNIC's history.

The strength of the first map was competence. Research networks and national registries were capable of solving a coordination problem. They could handle technical language, understand allocation practice and communicate with IANA and other registry actors. Their early role made APNIC operationally plausible.

The weakness was representational thinness. A research network can be a pioneer without being a proxy for a whole future market. A national registry can aggregate some local demand without representing every provider. A service provider can be a member without speaking for users or competitors. Early connectivity identifies where the Internet already existed. It does not identify where the Internet would later matter most.

This does not discredit the first map. It explains it. APNIC began from the parts of the region that could participate. The later governance test is whether the institution's access mechanisms expanded as the region's connectivity expanded.

The 2000 map expanded, but it was still institutional

By April 2000, APNIC's status report presented a larger institutional map: 396 members in 34 of 62 then-described economies. That is a substantial change from the pilot-end figure. It shows that the institution grew beyond a small founding circle and reached a much wider geography before the end of the first mass-Internet decade.

The number should be given its due. Three hundred ninety-six members is not a symbolic membership base. Thirty-four of 62 described economies is broad institutional coverage. It supports the view that APNIC had become a regional service body rather than a narrow research-network project. It also gives a useful comparison point near the beginning of wider commercial and public Internet growth.

But the figure remains institutional. It counts APNIC members and described economies with membership presence. It does not count all networks, all resource holders, all meeting entities, all voters, all request outcomes, all users or all residents. It does not tell us whether the 34 economies were evenly represented. It does not tell us whether large later user populations were already visible in APNIC governance. It does not tell us whether members in one economy had the same service quality, language access or meeting access as members in another.

The 62-economy denominator also requires care. APNIC currently describes a 56-economy region. The change from 62 to 56 may reflect differences in counting, boundary definition, economy naming or institutional presentation. The record used here does not supply a reconciliation. Therefore, the 2000 34-of-62 figure and the current 56-economy statement should not be joined as if they were a continuous series. They are separate boundary definitions that need explanation before comparison.

This is a common problem in institutional history. The institution's reported geography changes in form as the institution matures. Later readers are tempted to compare figures directly because the name APNIC remains constant. But denominators are part of the claim. If the number of described economies changes, the measurement frame changes too.

The safest conclusion is that the institutional map expanded significantly by 2000, while the public record here does not prove that expansion matched population, Internet-use growth or operator representation economy by economy.

The 62-to-56 problem is not clerical

The difference between the April 2000 report's 62 described economies and the current 56-economy regional statement may look like a clerical issue. It should not be treated that way without evidence. Boundary denominators are governance claims. They define the population of possible comparison, the economies counted as inside the service area, and the denominator against which coverage is advertised.

There are several possible explanations. Some entries may have been counted separately in 2000 and grouped later. Some names may have changed. Some territories may have been described differently for administrative reasons. Some current boundary language may use a narrower institutional convention than the older status report. The fixed evidence here does not show which explanation is correct.

That uncertainty affects every percentage or coverage claim. Thirty-four of 62 described economies in 2000 is about 2000's institutional frame. A current 56-economy statement is about today's frame. If a researcher silently treats 62 and 56 as the same denominator, coverage can be overstated or understated. If an institution wants to compare growth over time, it needs a concordance: each 2000 entry, each current entry, the reason for any change, and whether the change is naming, grouping, scope or institutional policy.

The same issue matters for accountability. If an economy disappears from a denominator because of a counting change, its operators do not necessarily disappear from the service problem. If an economy is grouped with another, service access and participation may still be uneven inside the group. If an economy is newly included, early absence cannot be held against APNIC's founding representation without dating the inclusion. Boundary clarity is therefore not an archival luxury. It is a condition for honest measurement.

The remedy is simple: publish a boundary concordance table for historical APNIC economy counts. It should include the 1994 pilot economies, the 2000 62-economy frame, the current 56-economy frame, name changes, grouping changes, service-status changes and the effective dates of each classification. Without that table, scale analysis can identify the problem but cannot compute clean historical coverage.

The current boundary is an endpoint claim, not a time series

APNIC's current 56-economy regional statement is important because it describes the present service boundary. It shows the scale of the region for which APNIC now claims operational responsibility. It also creates a temptation to read the current boundary backward into the early 1990s.

That would be a mistake. A current boundary statement can identify the endpoint. It cannot establish the historical reason for each inclusion or exclusion. It cannot show when each economy became practically connected to APNIC's service. It cannot show when operators in each economy became members, resource holders, voters, meeting attendees or policy entities. It cannot reconcile the 62 economies described in the April 2000 status report with today's 56-economy statement.

The current boundary matters because it frames the scale of today's institutional accountability. If APNIC serves 56 economies now, then access, language, time zone, fee, meeting and service questions should be tested across that boundary. But the current boundary cannot be used to say that early APNIC represented all 56 economies in the founding period. Nor can it be used to make a simple trend line from 12 economies in 1994 to 34 of 62 in 2000 to 56 today.

The right use is diagnostic. The current statement tells researchers what boundary APNIC now treats as its region. The early figures tell researchers what APNIC's institutional constituency looked like at two earlier moments. The gap between them identifies what must be measured: member growth, resource-holder distribution, voter participation, meeting attendance, request outcomes, fees and service quality by economy over time.

Until that series is built, scale claims should remain modest. APNIC clearly grew from a small early membership into a broad regional registry. The available figures do not prove that growth tracked population, Internet use or affected-network distribution.

The unique-ledger case comes first

Any serious critique of demographic mismatch has to begin with the registry's actual function. APNIC is not a population allocator. Its core role is to maintain and administer a unique number-resource ledger for network operators. Duplicate allocation would damage routing, trust and operational coordination. Regional boundaries are partly an efficiency mechanism: they reduce confusion over which registry handles which requests, records and policies.

That unique-ledger case is strong. It explains why a regional registry does not need to mirror population. It explains why a small number of operators in a small economy may create more immediate registry work than a large population with low network deployment. It explains why resource requests, not residents, drive daily service. It also explains why national registries, remote participation and local-language support can improve access without redrawing the ledger by demographic weight.

This is why a population mismatch should not become an argument for allocating addresses by population. Population may signal future demand, public importance and governance risk. It does not tell the registry how many unique number resources a network can justify. Allocating by population would confuse public scale with operational need.

The same caution applies to Internet-user percentages. A high percentage of individuals using the Internet signals dependence on connectivity. It does not directly identify the operators requesting resources from APNIC, the size of their networks, or their participation in APNIC governance. Users are downstream of resource administration. They are not the ordinary unit of registry allocation.

The governance concern is therefore not "large populations deserve more addresses." It is "large and growing affected publics deserve assurance that the registry's operator-facing institutions remain accessible, fair and responsive across the region." That is a different and better question.

The unique-ledger case protects APNIC from crude demographic arithmetic. It does not protect APNIC from access testing. A registry can allocate by network need while still measuring whether early connected economies retain disproportionate influence through meetings, voting, staff familiarity, language, travel, fees or service speed.

Internet-use data are context, not membership

World Bank and ITU Internet-use evidence is valuable because it places the institutional record against the region's later social scale. If an economy moves from limited early connectivity to broad public Internet use, the registry's public importance rises even if its direct member base remains operator-facing. A registry that was once relevant mainly to research networks and early providers becomes indirectly relevant to households, schools, businesses, mobile users and public institutions.

That does not make Internet users APNIC members. It does not make a user percentage a voting formula. It does not show how many providers need address resources. It does not prove that a particular economy's operators were underrepresented. It does something more modest and still important: it identifies the public dependency that makes institutional access worth testing.

The danger is to use user data emotionally rather than analytically. A large population or fast-growing Internet-user base can make early membership figures look morally small. That may be rhetorically tempting, but it is not enough. The right comparison is not residents versus addresses. The right comparison is later operator and service dependence versus the institution's access channels. If an economy's public connectivity grew rapidly, did its operators become APNIC members? Did they hold resources directly or through national registries? Did they attend meetings? Did they vote? Did they submit requests successfully?

Did they experience language or time-zone barriers? Did they use complaint routes?

Internet-use data can also reveal the risk of retrospective complacency. If an early connected economy remains highly visible in APNIC governance while a later high-use economy remains institutionally quiet, the institution should not assume the quiet economy is satisfied. It may be served through an intermediary, or it may face barriers. Without economy-level APNIC data, the public cannot tell.

This is why the article treats World Bank and ITU evidence as independent context. It supplies the outer frame of adoption. It does not replace APNIC's own missing member, voter, request and resource-holder series. The external denominator tells researchers where to look; the institutional denominator must show what happened inside APNIC.

National registries and remote access can reduce, not erase, mismatch

The strongest counterargument to the scale critique is that APNIC's operator-facing design can use intermediaries. National Internet registries, local operator groups, remote participation and local-language material can lower the cost of access without requiring every operator to deal with APNIC in the same way. In a region as large and heterogeneous as Asia-Pacific, that is not a minor point. Intermediation may be necessary.

National registries can aggregate local knowledge. They can communicate in local languages, understand domestic market structure and help operators that would not otherwise attend regional meetings. Remote participation can reduce travel burdens. Local-language support can make rules less dependent on English fluency. Smaller meetings and local training can bring registry knowledge to places that were not early research-network hubs.

But each mechanism needs evidence. A national registry can reduce access cost if it is accountable to the operators it serves. It can also become a second gatekeeping layer if fees, portability, voting or contract rights are unclear. Remote participation can broaden access if it allows meaningful intervention rather than passive observation. Local-language support can help if translated material arrives before decisions are effectively settled. Training can build capacity if it connects to actual membership and request rights.

For item 034, these mechanisms are counterevidence rather than proof. They show why demographic mismatch need not invalidate APNIC's boundary. They also show what to measure. If NIRs, remote participation and language support worked well, then economies with later connectivity growth should show improving membership, request success, meeting participation or policy input over time. If they did not, early geography may have persisted more strongly.

The fixed evidence here does not supply those outcome series. It therefore cannot say that intermediation solved the scale problem. It can say that intermediation is the plausible institutional route by which APNIC could reduce mismatch while preserving a unique regional ledger. The remedy is to document performance, not to assume it.

Persistence is a hypothesis, not a finding

The article's central hypothesis is that early connected-network geography can persist as institutional weight. The mechanism is plausible. Early connected institutions become early members. Early members learn meeting rules. They accumulate resource holdings. They know staff and procedures. They can attend policy sessions. They can nominate candidates. They can shape norms before later operators arrive. Their economies may appear more institutionally present even after user geography changes.

But plausibility is not proof. The fixed record here does not supply a consistent 1993-2010 series for APNIC members, voters, meeting attendees and resource holders by economy. It does not supply application success rates, service latency, fee burdens or appeal outcomes by economy. It does not prove that broadband inequality changed any specific APNIC decision.

The correct statement is conditional. If early connected economies retained higher membership density, higher voter turnout, greater meeting attendance, larger resource holdings or better service access after later user growth elsewhere, then early connectivity would have persisted as institutional weight. If later growth produced broad participation, balanced service outcomes and meaningful remote access, then the persistence hypothesis would be weakened.

The counterfactual is also important. Suppose APNIC had begun from population weighting rather than reachable networks. That might have looked more representative on paper, but it could have failed to solve the unique-ledger problem because many high-population economies had limited early operator capacity. The institution needed the networks that could actually coordinate. A purely demographic founding map could have been less operationally competent.

Disconfirming evidence would include economy-year data showing that membership, voting, meeting participation, resource holdings, application success and service quality expanded in line with later operator growth across the region. It would also include evidence that remote participation, national registry channels and local-language services reduced access differences. Without those data, the persistence hypothesis should guide measurement rather than harden into accusation.

This discipline matters. It keeps the article from turning into a digital-divide slogan. The existence of later inequality does not prove institutional bias. It identifies the variables that a serious institution should measure.

A measurement design for the missing panel

The future test should be an economy-year panel from 1993 to 2010. Each row would be one economy in one year. The unit is economy-year. The dependent variables would not be population. They would be APNIC institutional outcomes.

The panel should include: APNIC members by economy; resource holders by economy; address and ASN holdings by economy; new requests by economy; approved requests by economy; rejected or returned requests by economy; average service time by economy; fee burden by member size; meeting attendance by economy; remote participation where available; voter turnout by economy; candidates by economy; Executive Council representation; policy proposal authorship; national Internet registry channel availability; language support; appeals or formal complaints; and service incidents.

Independent context variables would include population, individuals using the Internet as a percentage of population, broadband adoption context where available, international connectivity indicators, and presence of research networks or national registries. Those variables should be treated as context, not as automatic entitlement.

The key ratios would be explicit. Members per known operator where operator data exist. Resource holders per Internet-user population where only broad context exists. Meeting attendees per member. Voters per eligible member. Average service time per request. Fee burden as a share of relevant member revenue or local cost proxy where data allow. The panel would also flag whether an economy appears in the 2000 62-economy frame and the current 56-economy frame, rather than silently merging the two.

The design could test whether early connected economies retained disproportionate institutional presence after controlling for operator counts and resource demand. It could test whether later high-user economies had lower meeting participation or slower service. It could test whether national registry channels changed access. It could test whether remote participation reduced travel bias.

What it still could not prove on its own is consent. A balanced panel would show better institutional access. It would not show that users authorised APNIC. A skewed panel would show a governance risk. It would not prove that any specific decision was caused by early connectivity. Causation would require decision-level evidence, meeting records, proposal histories and counterfactual analysis.

The panel is therefore a measurement tool, not a verdict machine.

It should also be published in a form that lets outside readers reproduce comparisons. A chart without the underlying economy-year rows would repeat the same problem in prettier form. Each value should identify the unit, the denominator, the data source, the date and the reason an entry is missing. Missing early Internet-use data should not be backfilled by assumption. Missing APNIC participation data should not be hidden by regional averages. The point is not to create a perfect dataset; it is to stop scale claims from depending on invisible denominator choices.

The access remedy is not population allocation

The remedy for scale mismatch is not to assign addresses or votes by population. That would confuse the registry's operational function. The remedy is to lower the cost of institutional access across the actual service boundary.

First, APNIC should publish economy-level participation and service metrics in a stable format: members, resource holders, requests, response times, meeting attendance, remote participation, voting, proposal authorship and formal complaints. The point is not to shame small economies or large ones. It is to see whether the institution's membership machinery reflects the operator geography it serves.

Second, APNIC should explain boundary denominators. If one record uses 62 described economies and the current statement uses 56, the institution should publish a reconciliation table. Which entries changed? Were they naming differences, scope changes, political definitions or administrative presentation? Without that table, historical comparison remains fragile.

Third, access support should follow measured gaps. If some economies have members but little meeting participation, remote formats and time-zone design may matter. If some economies have operators but few members, membership cost or national registry structure may matter. If service latency varies, staffing and language support may matter. If appeal or complaint use is absent, the remedy may be too costly or poorly understood.

Fourth, the institution should distinguish user-facing public importance from operator-facing allocation rules. It can acknowledge that millions of users depend on networks affected by registry administration without pretending those users are members or address applicants. That honesty strengthens the institution because it shows the public consequence without distorting the operational rule.

The access remedy is therefore empirical: measure who participates, who receives service, who can complain and who remains invisible. Then adjust access channels without making population the allocation rule.

The finding: scale tests access, not address entitlement

The ranked denominator finding is this.

The strongest evidence is institutional: APNIC reports 27 members in 12 economies in 1994 and 396 members in 34 of 62 described economies in April 2000. These figures prove growth in recorded membership and economy coverage under APNIC's own institutional frame.

The moderate evidence is contextual: research-network histories identify early connected institutions, and World Bank or ITU data can show later Internet-use and broadband context. These data help describe the gap between first connected networks and later public dependence. They do not prove APNIC representation or resource demand.

The weakest evidence is causal: the current record does not prove that demographic or broadband inequality changed a specific APNIC decision. It also does not prove that early connected economies retained disproportionate weight after 2000. That remains a testable hypothesis.

The access remedy is to publish stable economy-year participation and service metrics, reconcile the 62 and 56 economy frames, and target remote, language, fee and complaint support where the data show gaps. The economy-year measurement test should track members, resource holders, requests, outcomes, service time, meetings, votes, proposals, national registry channels and user-context variables by economy from 1993 to 2010. The archival test is a complete series of APNIC membership, voter, attendance, request, resource-holder and boundary-definition records with enough metadata to compare economies without changing denominators silently.

Until those records are public, the precise conclusion is narrow. APNIC's early map followed the reachable networks of the pre-broadband region. That was operationally rational for a unique ledger. It was not proof that later population scale or Internet-use geography had been represented from the beginning. The mismatch does not justify allocating numbers by population. It does justify measuring whether the institution's access routes grew as the region it served became much larger than the room in which its earliest entities could gather.