Summary

  • ARIN’s 13-site criterion gives a qualifying end-user network another route to an initial IPv6 assignment: at least 13 active sites in a contiguous network, with operation intended within 12 months. It recognizes that a dispersed organization can face substantial renumbering costs even when each location has few users or subnets.
  • David Huberman was one of two Advisory Council shepherds of ARIN-2015-1, not its originator. The proposal’s record names Gary T. Giesen as originator and Scott Leibrand and Huberman as shepherds. The rule became part of policy, but the public record reviewed here does not show how many organizations used it or whether it caused broader IPv6 deployment.

A count that tries to see a cost

Imagine an organization with fifty stores, depots or offices connected through one private IP virtual private network. Each site has a handful of staff, a checkout terminal or two, a small voice network and a data network. Across the company, the total number of addresses in use may be modest. The network may have only two subnets at each location. It may rely on one carrier and have no operational reason to multi-home every location to the public Internet. Yet changing the upstream arrangement, the addressing plan or the enterprise’s internal network can reach dozens of places and the people who depend on them.

That was the kind of organization ARIN’s 2015 proposal used to explain a gap in the existing tests for an initial end-user IPv6 assignment. The example in the proposal was explicitly hypothetical: fifty locations on an IPVPN, an average of ten staff and twenty devices per site, and two subnets per site. Its purpose was to show why address, subnet and user counts can move independently of the number of operational places that must be renumbered. It was not a named customer case, a survey, or evidence that a particular company had been denied space.

The policy question was narrower than whether IPv6 is useful or whether every enterprise should receive a direct block. It asked what evidence could qualify an end-user network whose costs of renumbering might be high even when its address use did not meet the numerical thresholds already in policy. An organization might receive provider-assigned addresses from an internet service provider or another local internet registry. Those addresses can be suitable for many networks, but a move between providers can require a change in the organization’s addressing.

The proposal argued that, for some dispersed networks, the transition cost was poorly represented by a simple count of active IPv6 addresses or /64 subnets.

A site-count threshold cannot measure every cost of a network change. It can, however, establish a public rule that administrators and applicants can discuss. That move from an unstructured claim about “large renumbering costs” to a count tied to a network footprint is the central design choice in ARIN-2015-1. It is also the reason the number thirteen deserves more attention than the number alone suggests: the rule treats geography and topology as evidence about a cost that is otherwise difficult to observe.

A career across the registry and the operator’s problem

David Huberman’s professional history places him at several points in the address-management system. In a 2015 ARIN voter guide, he described his career as ARIN hostmaster from 1999 to 2000, global IP addressing manager at Global Crossing from 2000 to 2002, and hostmaster at ARIN again from 2003 to 2013. He said he moved to Microsoft in 2013 as Principal, Global IP Addressing, and had joined ARIN’s Advisory Council in January 2015. These dates are Huberman’s candidate-statement account, not an independently audited résumé. ARIN’s 2015 voter guide

That path crossed the administrative and operational sides of number resources. A registry records and administers allocations under policy. A large network operator plans address use, connects locations and peers, and decides which prefixes it will accept or announce. The rules that sit between those functions determine who can request space, what evidence is expected, and how large a block may be justified. Huberman’s work history therefore provides context for his participation in policy discussions, but it does not establish what he privately believed about ARIN-2015-1 or that his employers shaped its text.

His 2015 statement described the Advisory Council as an important channel for policy improvement and stressed his long experience on both the staff and community sides. It also included his view that ARIN’s role as steward of number resources should remain generally stable after IPv4 exhaustion. Those were arguments made by a candidate seeking an elected position. They are useful for understanding how he presented his experience to ARIN’s members; they should not be treated as proof that all voters shared that view or that any one candidate determined the Council’s policy direction.

The later stages of his career offer a separate endpoint. ICANN’s staff biography says Huberman joined ICANN in 2018 after ten years at ARIN and identifies him as Technical Engagement Director for North America and global standard development organizations. In April 2026, ICANN published an Office of the Chief Technology Officer paper on ICANN and the IETF under his name. ICANN cautions that profile details are not independently verified by the organization, so the biography is best read as the institution’s published account. The report is evidence that he continues to write about technical coordination; it is not evidence about ARIN’s allocation outcomes. ICANN staff biography · ICANN OCTO-043

Direct assignment is a policy outcome, not a network outcome

The term “direct” needs care. ARIN’s rule concerns an initial IPv6 assignment to an end user: an organization that operates a network for its own use rather than an internet service provider that reassigns addresses to customers. It does not mean that every end user must obtain addresses directly from ARIN. An ISP or another local internet registry can provide address space, and the policy lists more than one route for an end user to justify an initial assignment. The 13-site criterion is one alternative among six in the current Number Resource Policy Manual, not the test every applicant must pass. ARIN Number Resource Policy Manual, §6.5.8.1

The other criteria are important because they define the actual scope of the new one. An organization may qualify through a previously justified ARIN IPv4 end-user assignment, present or imminent IPv6 multi-homing with a valid global autonomous system number, active use of at least 2,000 IPv6 addresses within 12 months, active use of at least 200 /64 subnets within 12 months, a reasonable technical justification for why addresses from an ISP or local internet registry are unsuitable, or the contiguous-network test of at least 13 active sites. Each is a distinct policy path.

An organization with fewer than 13 sites is not barred from an initial IPv6 assignment if it meets another criterion.

The initial-assignment rule also specifies an intended operational horizon: the addresses are to begin operational use within 12 months. That statement matters because a registry is not simply handing an organization a symbolic entitlement disconnected from a network plan. An applicant needs a documented reason to request the resource under policy. But eligibility is still only the first step.

It does not establish that ARIN has approved a particular request, that the initial block will be a specific size, that an upstream provider will carry it, that the organization will announce it in BGP, or that devices and services have completed a migration.

This separation is often lost when a policy threshold is summarized as “13 sites get IPv6.” The more exact description is that an end user can use the 13-active-site route to justify an initial assignment if the network is contiguous and the sites are within the stated planning horizon. The policy makes an additional eligibility argument available. It does not promise a route advertisement, worldwide reachability, a service-level outcome or a successful deployment. Those are later decisions involving the applicant, ARIN, network operators and vendors.

The distinction is commercially material. An organization deciding whether to use provider-assigned space or seek its own assignment must consider more than eligibility. Direct assignment may make address-number-resource continuity less dependent on a particular provider relationship, but it also creates operational duties: planning the prefix, maintaining registration information, arranging routing, operating DNS and coordinating transition work. The policy does not price those tasks or make them disappear. It supplies one way to explain why the address resource may be justified.

Why count sites when addresses are already countable?

The earlier tests were built around address use, subnet use, multi-homing or a documented exception. Those measures address different situations. A large campus or a high-volume network can demonstrate substantial address use. A multi-homed organization has an explicit routing relationship that may justify a portable block. A technical justification can explain special circumstances that do not fit the other thresholds. The proposal’s claim was that these routes did not give every geographically dispersed, low-count organization a predictable way to demonstrate its cost of changing providers.

A location count is an imperfect proxy, but it measures something the other numbers may not. A company can have many small locations with few users, devices and subnets at each one. In that case, an address threshold that sums devices or subnets may understate how many operational sites a new prefix or renumbering plan has to reach. The burden can include coordinating local equipment, service windows, applications, vendors and staff. These are plausible mechanisms for added cost, not quantities measured by the proposal. The public documents do not supply average outage time per site or a dollar estimate for the illustrative network.

The policy response was not to discard address or subnet thresholds. It added another route alongside them. That design preserves several ways to qualify and avoids making a single organizational shape the assumed norm. One large site might rely on address count or a technical justification. A service provider might use a different allocation policy. A dispersed end user could make a site-count case. The proposal’s example was deliberately a network whose low subnet and user counts did not capture its geographic footprint.

In the archived draft, the stated minimum was not 13 arbitrary offices located anywhere. The network had to be contiguous, its sites had to be active within 12 months, and the proposal tied the threshold to the existing number of sites associated with a /40 initial assignment. The explanation described a balance: reducing the applicant’s expected renumbering cost while acknowledging the burden of carrying an additional Internet prefix. Requiring contiguity was meant to make aggregation more feasible. A single aggregate does not eliminate the cost of carrying a route, but it constrains how the resource might be represented in routing.

That matters because a private organization’s continuity problem and the public routing system’s carrying cost do not fall on the same party. If the rules privileged only applicant convenience, they could understate the shared cost of a new prefix. If they privileged only aggregate efficiency, they could ignore costs that appear at the edge of a large provider network but are not visible in an address count. The 13-site rule tried to make both sides explicit in the same eligibility clause. It did not prove that the two costs are equal in each case, and it did not establish an empirical optimum at thirteen.

Its rationale was a policy calibration anchored to an existing threshold.

Thirteen sites, a /40, and the separate question of size

The number is easiest to misunderstand when the eligibility threshold and the resulting address-block size are collapsed. The archived proposal says 13 sites were selected with reference to the existing /40 initial-assignment site rule. That connection gives the number an internal policy logic: it reuses a visible boundary in the manual rather than inventing a number detached from the way ARIN already thought about site counts and aggregation. The current manual still presents 13 active sites in the eligibility section and separately explains how initial assignment size depends on the number of justified sites. ARIN-prop-215 archive · Current NRPM, §§6.5.8.1–6.5.8.2

ARIN’s current request guide says an organization that meets at least one initial end-user criterion is eligible for a minimum initial /48. If it requests a larger allocation, it must document the extra need. The guide lists 13 to 192 justified sites as an example corresponding to a /40, with larger site ranges tied to larger blocks. The manual’s size section also says the assignment is evaluated against site counts and subnet plans for extra-large sites. Thus the 13-site eligibility rule is not a guarantee that every applicant receives a /40, nor a stand-alone size calculator. It provides a way to satisfy an initial-eligibility criterion; requested size remains a distinct decision with its own evidence. ARIN resource request guide

That distinction is not semantic housekeeping. A /48 is the minimum initial assignment described in the current guide. A request for more room has to show why it is needed. If readers hear “13 sites means /40,” they may mistake an example of size guidance for an automatic entitlement. If they hear “13 sites are required for direct IPv6,” they may miss the other five eligibility routes. Both errors turn a multi-part policy into a slogan and obscure where ARIN’s discretion and the applicant’s documentation still matter.

The original text also included a 12-month horizon for sites that would be operational, rather than treating a future corporate site plan as sufficient indefinitely. Current policy language expresses intent for addresses to begin operational use within 12 months and uses the same horizon in the site criterion. This makes the timeline part of the evidence, not only the count. A firm with 13 sites today but no plan for address use would not satisfy the criterion merely by pointing to its org chart; an applicant must establish the network and operational case under ARIN’s rules.

The prefix-size connection reveals a tradeoff rather than a formula that can be optimized on paper. If the threshold is too low relative to the expected carrying cost, the policy may make additional prefixes available to organizations whose operational case is less compelling. If it is too high, some networks with meaningful provider-change costs may not have a clear path. The proposal’s approach was incremental: add a new objective route that fits a known site boundary, preserve the rest of the criteria and keep the registry’s documentation process in place.

That choice let the community debate the exact change without reopening every assignment rule at once.

What counts as a site is a network question

A site sounds like a simple unit until a network is distributed across a campus, a set of buildings, a retail chain or a collection of facilities connected through different kinds of transport. ARIN’s current manual defines a site as a discrete location that is part of an organization’s network. It says a campus with multiple buildings may be treated as one site or several depending on how its network infrastructure is implemented. To count the campus as multiple sites, the organization must submit reasonable technical documentation showing an implementation equivalent to multiple sites. The same manual says an organization may request up to a /48 for each site and sites expected to be operational within 12 months. Current NRPM, §6.5.8.2.1

This definition moves the test away from mere real-estate enumeration. A firm cannot automatically transform every address, floor, store room or legal entity into a separate “site.” The network architecture matters. Multiple buildings on one campus might be one network location, or their design might support separate network sites if the organization can document the distinction. The criterion therefore joins an objective count with technical judgment. It tells an applicant what kind of claim can be made, but it does not remove the need to describe how the organization’s network works.

There are several values in that arrangement. Documentation can prevent site count from becoming a loophole in which labels substitute for operational networks. At the same time, a decision process that depends on documentation can impose costs on organizations with smaller policy staffs. The current sources establish the documentation requirement, but they do not report how often an application was challenged, how much staff time was needed, or whether small organizations faced a higher rejection rate. Those are reasonable monitoring questions, not conclusions that can be drawn from the wording alone.

The term “active” carries a similar issue. The current criterion attaches activity to a 12-month horizon. That requirement makes the threshold forward-looking: the organization must have an intended operational use, not only a list of potential facilities. But the documents reviewed here do not furnish a public dataset about how “active” is evidenced across different industries or how ARIN handled edge cases. An analyst should not claim that the manual contains a universal measurement method beyond the rules it actually publishes.

The significance of the site definition is that it makes the threshold both countable and contestable. Countable evidence gives applicants a standard to organize around. Contestability lets the registry ask whether the count corresponds to network structure. In return, the applicant must expose enough of its internal network plan to justify the claim. The policy does not decide the internal design for the organization; it uses the organization’s own network design as evidence for a resource decision.

A shepherd is not an author or an approving authority

The commission turns on Huberman’s role, so the public record must distinguish several jobs that can be conflated in short descriptions. ARIN-prop-215 names Gary T. Giesen as the proposal originator. The later ARIN-2015-1 archive lists Scott Leibrand and David Huberman as Advisory Council shepherds. In the ARIN 35 public policy meeting transcript, Leibrand presents the draft and explains its rationale. ARIN 36 also identifies the shepherds and describes the policy as an additional route for an end-user organization to qualify. The available sources therefore support calling Huberman a co-shepherd, not the proposal’s author or sole designer.

That is not a minor distinction. The proposal originator brings a problem statement and text into the policy development process. An Advisory Council shepherd helps move the policy through the Council’s work and public discussion. The Council can recommend an outcome, and the Board can ratify a recommended policy. Staff then implements the rule and updates guidance. The public record shows those distinct stages. It does not give any one participant unilateral authority to turn a proposal into policy.

A shepherd can matter without owning the idea. The title indicates a formal place in the process, not proof of the private contribution each person made to every wording change. Available meeting records document named shepherds and presenters, but they do not quantify who wrote which line, who persuaded a participant, or what would have happened without any individual. The appropriate profile claim is consequently bounded: Huberman was one of two people named to shepherd this proposal through ARIN’s Advisory Council. That role placed him inside the governance path; it did not make him the single decision-maker.

This boundary is especially important because the content concerns a rule that could affect applicants and the broader routing system. Overstating one person’s authorship would hide the role of Giesen, Leibrand, other Council members, meeting participants, the ARIN community, the Board and staff. It would also confuse expertise with authority. A technical professional may bring useful experience to a debate, but the policy gains legitimacy through its process, published reasoning and accountable decisions, not through a biography alone.

The documents show Leibrand introducing the draft at ARIN 35, with the room discussing whether the Council should continue work. The poll recorded 28 in favor and one against among 112 participants. That was a consultation signal to the Council, not a vote of all ARIN members or the final adoption decision. Later, on 18 June, the Advisory Council unanimously recommended the policy. The Board ratified it in December, and ARIN records implementation in February 2016. Each vote answers a different question at a different stage; combining them into a single “community approved it” sentence would erase the process.

From proposal to implemented rule

The proposal originated as ARIN-prop-215 in February 2015. It argued that the existing end-user criteria could exclude some organizations whose costs to renumber exceeded what address and subnet counts were designed to capture. The draft would add a contiguous network with at least 13 active sites within 12 months as another way to justify an initial assignment. The original proposal named Giesen as originator; the later AC shepherds were Leibrand and Huberman. The archived Recommended Draft Policy says the new criterion was intended to make administration fairer for dispersed or low-count organizations while preserving a reasonable threshold that supported hierarchical aggregation. ARIN-prop-215 · ARIN-2015-1 archive

At ARIN 35, Leibrand presented the proposal and described a dispersed user whose many locations could make an upstream change expensive even when the site-level network was small. The meeting’s poll asked whether the Advisory Council should continue work on the draft. It drew 28 in favor, one against, out of 112 participants. That signal helped answer whether further work should proceed. It was not a binding referendum and did not show that every participant who did not vote opposed the policy. A public meeting transcript captures the meeting’s discussion and poll; it is not a region-wide survey.

The draft also appeared at the Public Policy Consultation at NANOG 64 and at ARIN 36. The record gives the process a sequence: a proposed problem, an initial draft, public discussion, Advisory Council work, a recommended draft, Board ratification and implementation. On 18 June 2015, the Advisory Council voted unanimously to recommend ARIN-2015-1 for adoption. The Board ratified it on 10 December 2015. ARIN’s archive records implementation on 17 February 2016. These are independent steps; a committee recommendation is not the same as Board ratification, and ratification is not the same as an operational change by staff.

The sequence helps explain what a shepherd’s authority can and cannot mean. Shepherds are attached to the proposal while the Council develops and evaluates it. They may present it and help the Council consider revisions and feedback. But the decision passes through a structure that distributes responsibility. The community’s public record shows arguments and signals; the AC makes a recommendation; the Board ratifies; staff implements. No evidence here supports a narrative in which Huberman personally carried the policy from concept to outcome or persuaded the Board by himself.

That institutional structure matters for future research as much as for historical credit. If a policy later changes, an analyst should look at the proposal record, the meeting transcript, AC decisions, Board minutes, implementation date and current manual. A title or byline alone cannot reveal which version reached the rulebook or whether a criterion was preserved. ARIN-2015-1’s current wording can be checked directly against the 2015 archive; the archive, in turn, shows how a criterion that remains current passed through formal stages.

The 2014 proposal is a contrast, not a predecessor

Huberman’s record includes a separate policy episode that illustrates why authorship and shepherding should not be merged. In a February 2014 message to ARIN’s Public Policy Mailing List, he said he was the author of ARIN-2014-3, a proposal to remove the /24 minimum block size from IPv4 transfer policies. His argument was that network operators, rather than address policy, should decide what prefix sizes they would export to peers. That was his position in a community discussion, not a general principle adopted by ARIN and not a statement about the later IPv6 site threshold.

The archive lists Owen DeLong and Bill Darte as the AC shepherds of ARIN-2014-3. In June 2014, the AC abandoned the draft. The meeting record cites minimal mailing-list support, weaker support at ARIN 33 and a 22-to-1 signal for abandonment at a recent public policy consultation. The mailing-list explanation said that although there was some support for reducing the minimum block size, the Council did not see enough support to change the policy as proposed. The records therefore show Huberman as a self-identified author of that draft, while showing other people as shepherds and a community process that ended in abandonment. ARIN-2014-3 archive · Huberman’s mailing-list message · June 2014 AC meeting

The episode is useful because it demonstrates that a person can originate a proposal without being its AC shepherd, and a shepherd can be involved in a proposal without being its originator. It also shows that the policy system can reject or abandon an idea when community support is insufficient or the proposed change does not attract consensus. Those facts help interpret Huberman’s later position on ARIN-2015-1: his formal role differed, and the outcome differed. There is no evidence that the IPv4 transfer proposal caused the IPv6 eligibility rule, that the two were a coordinated program, or that one was a response to the other.

A later ARIN-2016-3 policy on simplified criteria for some small IPv4 transfers was shepherded by Huberman and Owen DeLong and implemented in 2017. It concerned a different section of policy and a different transfer mechanism. It can be mentioned as evidence that Huberman continued to participate in ARIN policy work, but it does not establish a chain from the abandoned 2014 proposal to the 2015 IPv6 rule. This article keeps that later example in the background because the 13-site eligibility test is the subject, not a survey of every policy he touched.

The narrower comparison also avoids reading a person’s career backward from a policy’s result. A proposal can express an idea that its originator supports; a shepherd can support or critique a text while helping the Council develop it; an institution can recommend, ratify or abandon it. Those actions leave different public traces. Personal intent is not always recoverable from meeting minutes, so the article describes only what the available sources show.

What the 2015 record can establish

The record is unusually clear about the rule’s design rationale. It specifies the 13-site test, notes the relationship to a /40 threshold, describes the contiguity condition, and gives a hypothetical multi-location example. It identifies the originator and shepherds. Meeting records show public presentation and a consultation signal. AC and Board documents show the recommendation and ratification; the archive shows implementation. The current NRPM confirms the criterion remains in force, while ARIN’s request guide explains the present relationship between eligibility, the minimum /48 and larger blocks justified by site counts.

Together, those sources establish that the rule was not a shortcut to universal direct IPv6. They establish that ARIN created one additional way for certain end users to justify an initial assignment, that the condition was tied to both an operational horizon and a contiguous network, and that policymakers explicitly considered applicant renumbering costs alongside prefix-carrying costs. They also show that a site is a network location, not automatically any legal address an organization chooses to count.

They do not reveal the number of applications that used the criterion, the approval rate, the typical size of assignments granted under it, how ARIN applied the active-site condition in difficult cases, or how many organizations started IPv6 afterward. They do not provide a before-and-after comparison of provider changes, renumbering incidents, BGP announcements, route-table growth or network uptime. The policy’s presence in the manual is proof of a rule, not proof of its use or impact.

That distinction matters for evaluating both praise and criticism. It would be unsupported to say the 13-site test accelerated IPv6 adoption, removed provider lock-in or saved end users a quantifiable amount. The proposal expressed a reason to expect a benefit, and the rule offered a route for applicants who could meet it. Whether it changed real choices is an empirical question. It would also be unsupported to call the criterion ineffective merely because this set of public documents lacks take-up statistics. Missing evidence is not evidence that applicants did not use the rule.

An evidence-driven evaluation would connect several datasets while respecting privacy and the limits of what each can show. An aggregate count of requests that cited the 13-site criterion could establish usage. Decisions and requested block sizes could show how the registry applied the rule and how often eligibility translated into a larger assignment. Publicly observable routing data might show whether assigned resources were announced, although it would not identify every internal deployment or prove why a route appeared.

Surveys or case studies could test whether organizations experienced lower renumbering costs, but would need a clear definition of cost and a credible comparison group. Longitudinal deployment data could address adoption; a policy archive alone cannot.

The rule today: a current test with separate downstream decisions

The current Number Resource Policy Manual retains the contiguous network criterion. Its wording calls for a minimum of 13 active sites within 12 months. The adjacent size section says an organization meeting at least one initial-assignment criterion is eligible for a minimum /48; larger requests are reviewed with supporting documentation. The manual defines a site as a discrete location in the organization’s network and addresses how a campus may count as one or multiple sites. The current request guide translates this into applicant-facing examples, including 13–192 justified sites as a /40 example for a larger initial request.

These sources make it possible to describe the present rule without assuming that every detail is the same as the 2015 proposal’s explanation.

For a network operator, the practical sequence is therefore not “count thirteen and receive a routed prefix.” It is closer to: identify the correct end-user policy category; choose an available eligibility path; document the network and intended use; justify size above the minimum; receive and register resources if approved; and then arrange routing, internal addressing and migration. Each stage can have its own actor, evidence and risk. A direct assignment may provide continuity in the address plan while still leaving the operator dependent on transit providers to announce or carry the prefix.

The policy cannot decide a commercial contract, guarantee upstream acceptance or operate the network.

That division of responsibility reflects a wider internet-number governance problem. A regional registry administers a pool under policy; an organization builds an internal network; an upstream provider decides what it will carry; the global routing system reflects many operators’ choices. A rule that qualifies one end user for a resource affects only one layer. It does not override other operators’ routing policies or make the Internet treat every announced prefix identically. The 2015 proposal’s emphasis on contiguity signals awareness of shared routing considerations, but it is not a measurement of actual global impact.

The 12-month horizon links the policy to planning, yet it also presents a future-facing question: what happens when sites change before a request is completed or after an assignment is made? Policy text and public guidance can describe when an applicant is eligible and how to justify size, but an organization still has to update the plan as sites open, close, merge or change their network design. Such ordinary business changes can alter an address plan.

The sources reviewed here do not document how ARIN handled every such scenario under this criterion, so claims about specific treatment should be checked against current staff guidance and the policy manual.

For this reason, the strongest current interpretation is modest. The rule gives the end user another documented basis to request direct IPv6 space. It gives the registry a site-based threshold to evaluate. It preserves a role for documentation and technical judgment. It does not force the organization to adopt IPv6, eliminate a provider relationship, assign a particular block size automatically, or make a route operational.

The monitoring questions that remain

If the rule is to be evaluated today, the first task is not to treat its survival in the manual as a success metric. The continued presence of a criterion shows that ARIN has not removed it from the current policy. It does not reveal how often it matters. A useful monitoring series would distinguish requests made under each eligibility criterion, the number of sites justified, the sizes requested and granted, and the interval from request to operational use. Counts should be aggregated in a way that avoids disclosing applicants’ internal network maps or commercial plans.

A second line of evidence concerns interpretation. Do applicants and ARIN staff use the same definition of “active site”? How frequently do multi-building campuses claim separate sites, and what documentation is accepted? Does contiguous network language create a predictable test for organizations using a single provider IPVPN, or does it exclude plausible cases where the organization’s topology is more complex? The current manual speaks to site definition and campus documentation, but those public provisions do not answer how consistent decisions are in practice.

An anonymized decision summary could reveal whether cases cluster around ambiguity in the policy text.

A third line concerns what happens after assignment. Is the block used within the stated planning horizon? Is it announced as one aggregate, delegated to multiple providers, or kept for internal-only use? Does the organization later renumber, and if so, for what reason? A public route collector can observe some external announcements, but it cannot see all internal use and cannot prove that the 13-site criterion caused a deployment decision. A registry’s request data can show application and assignment events but not necessarily successful migration. The two sets need to be combined carefully rather than used as substitutes.

Triggers for renewed policy discussion could include a sustained increase in requests made under this criterion, repeated interpretation disputes over site counts, a material change to the size rules that breaks the original alignment, evidence that the contiguous-network condition does not produce the expected aggregation, or a measurable pattern of applicants whose renumbering costs are not represented by the current paths. Conversely, little use over several years might invite a closer examination of whether the route is redundant with another criterion, but low use alone would not show that the rule lacks value.

It may be designed as an option for a narrow class.

There are also changes outside ARIN’s policy text that affect the applicant’s calculus. Provider-assigned IPv6, multi-provider network architectures, network automation, asset management and the cost of maintaining a portable address plan can all alter the tradeoff. A threshold introduced in 2015 should therefore be read against current operating practice, not frozen as a timeless answer. Yet any revision should preserve the distinction between site eligibility, size justification and routing behavior; changing one does not automatically resolve the others.

Conclusion: one additional lens on a dispersed network

ARIN-2015-1 did not make geography the only measure of an end-user network. It added a specific one: at least thirteen active sites, within twelve months and in a contiguous network, alongside other eligibility routes. The proposal’s logic was that a collection of low-count sites could impose a real renumbering burden that address and subnet totals did not capture. The same text recognized that an additional prefix has a shared carrying cost and used the existing /40 site threshold as a reference point.

Huberman’s documented role was bounded but meaningful. He was one of the Advisory Council shepherds who helped bring the proposal through a formal policy process. Gary T. Giesen originated it; Scott Leibrand also shepherded it and presented it in public meetings; the AC recommended it, the Board ratified it, and staff implemented it. That division of credit says more about the institution than a single-author narrative would. It shows that policy can originate in the community and pass through technical review, public discussion, committee recommendation and Board action before becoming an operational rule.

The rule’s continued place in the current manual is evidence of institutional continuity. Its actual use and effects remain unmeasured in the sources considered here. The distinction is not a reason to discount its design; it is a reason to ask for the right evidence. A threshold can be clear and still require judgment. A direct assignment can improve address-number continuity and still leave routing and deployment to other actors. Thirteen can function as an auditable minimum without being a universal proxy for every organization’s cost.

The most durable insight is that infrastructure policy makes choices about what it can count. ARIN already counted addresses, subnets and multi-homing. The 13-site criterion added another observable: how far a network is distributed across the places where it must operate. In the rule, that count does not replace technical documentation or downstream decision-making. It makes one previously less visible form of cost legible enough to enter the policy discussion.

Sources