Summary

  • LACNIC dual-stack cost incidence measures the fully loaded annual parallel-stack cost per active customer or revenue-critical application.
  • The bill moves through support queues, vendor and CPE gaps, security and observability duplication, wholesale terms, outages and product segmentation rather than one transition budget.
  • Transparent incidence and portable identity preserve operator choice; Number Resource Society offers the future-facing coordination layer for holders rather than a new mandate over deployment.

The incident ledger begins inside the wholesale bundle

The useful dual-stack story in Latin America and the Caribbean does not begin with a protocol diagram. It begins with a cost reconstruction after a service incident. A retail provider has business customers complaining that payment terminals fail intermittently, remote-access sessions drop, a hotel booking system behaves differently after a router replacement, and a municipal office can reach some cloud services but not the legacy supplier portal that closes its daily work. The network operations centre can show traffic passing. The wholesale carrier can show route acceptance. The device vendor can point to a firmware table. The managed firewall supplier can show that its policy matched one family better than the other. The cloud platform can show the extra public-address feature that kept one application visible. No single party has a line item called "dual-stack cost". The incident has already scattered the bill.

That scatter is the core economics. The LACNIC-region operator is not deciding whether IPv6 exists, whether IPv4 is scarce, or whether coexistence has technical merit. Coexistence is already part of the operating environment. The question is who bears the fully loaded annual cost of keeping IPv4 compatibility and IPv6 reachability reliable for the particular customer, site or application that would lose revenue if either path failed. Lu Heng's argument that "IPv6 transition" often functions as a permanent dual-stack tax is deliberately sharp, but the mechanism is plain: the second stack arrives before the first one can be retired, so operators pay for two assurance surfaces rather than one (heng.lu).

The LACNIC setting makes the allocation problem sharper because many services are sold through layered commercial bundles. A local access provider may buy upstream capacity, address support and route evidence from a wholesaler, sell a business plan to a shop, clinic, hotel or public office, outsource part of managed security to an integrator, and rely on cloud or payment providers whose identity assumptions were built elsewhere. The customer sees a single service. The cost sits across wholesale minimums, public-address inventory, device depreciation, first-line support, vendor escalation, cloud add-ons, customer credits and management time.

The right unit is not a generic "IPv6 programme" budget. It is the annual coexistence cost per active business customer, per public-facing service, or per revenue-critical application. That unit includes public IPv4 inventory or leasing, IPv6-capable equipment, monitoring, firewall parity, support scripts, route evidence, reverse-DNS continuity where customers rely on it, security and abuse handling, vendor workarounds, emergency changes and the expected cost of outage recovery. It also includes the cost of not spending: longer support queues, failed renewals, avoidable credits and customers buying weaker products because the provider cannot explain what assurance costs.

The incident ledger is useful because it refuses to let each party stop at its own contractual defence. The wholesaler supplied packets, the retailer owned the customer, the CPE vendor shipped a device, the firewall supplier supported a rule set, the cloud platform sold a feature, and the payment processor kept an old allowlist. All may be locally defensible. Together they create an unpriced system. Cost incidence begins when finance asks which party had the power to reduce the ambiguity, which customer benefited from the assurance, and which contract should recover the cost next time.

The reconstruction should be deliberately prosaic. It should list overtime, call handling, engineer escalation, emergency vendor time, customer credits, additional public-address products, temporary routing, replacement devices, delayed installations, churn risk and the managerial hours spent making suppliers agree on what happened. The point is not to invent precision where the records are poor. It is to prevent the cost from remaining invisible merely because it was booked under many ordinary headings. A provider that cannot reconstruct the incident cost cannot price the next service bundle; it can only hope the next fault is cheaper.

Incidence starts with the application, not the protocol

The first accounting error is to treat an address family as the cost object. IPv4 and IPv6 do not pay bills. Customers, applications and contracts do. A residential broadband line, a hotel reservation platform, a clinic's remote-support link, a customs broker's portal, a call-centre VPN, a municipal payment service and a wholesale retail handoff all consume coexistence differently. The same access network can serve them, but the assurance burden is not the same.

For a household plan, ordinary reachability may be enough if the customer has no inbound service, no stable public identity requirement and no revenue loss from an occasional application edge case. For a small firm, the same default can be inadequate. A shop may need card-terminal reliability, camera access, supplier portals, cloud accounting and predictable source identity. A clinic may need vendor support and patient-administration systems to recognise the network. A hotel may depend on booking platforms, payment gateways and guest Wi-Fi at the same time. The difference is not bandwidth. It is the cost of being recognised by counterparties.

That is why the application has to be the denominator. If one application requires dedicated IPv4 egress, tested IPv6 reachability, managed firewall parity, outage evidence and supplier coordination, its annual cost should not be hidden in the base price for every subscriber. If another application can rely on competent default access, it should not pay as if it were a bank integration. Product economics becomes fairer when the cost object is the customer function that creates the cost.

Lu Heng's agency-problem analysis is useful here because technology decisions are often promoted by actors that do not carry the operating company's cash-flow risk (heng.lu). Engineers may prefer architectural cleanliness, vendors may prefer refresh cycles, platforms may prefer priced exception menus, and institutions may prefer adoption language. The operator faces churn, credits, support labour and capex timing. A cost unit tied to the customer application forces the argument back to the balance sheet.

It also prevents false equality. A provider in Sao Paulo, a Caribbean hospitality supplier, a Central American enterprise integrator and an Andean fixed-wireless operator cannot apply the same regional average. Their customer mix, currency exposure, upstream choices, device fleets and support labour differ. The method can be common while the number is local: identify the revenue that depends on coexistence, list the duplicated assurances needed to protect it, and decide whether the customer, retailer, wholesaler, integrator or shareholder carries each part.

The application lens also separates this topic from neighbouring LACNIC arguments. Growth-pressure economics asks whether new demand can be matched with deployable identity quickly enough. Transition political economy asks why the final IPv4 exit remains non-exercisable. Dual-stack cost incidence assumes the old and new systems are both present and asks who pays for making the combined service believable today. That is a narrower, more contractual question.

It also changes the internal conversation. A network team may describe a "dual-stack customer" too broadly, as if the same label covers every household and every enterprise circuit. Finance should break that label into use cases. Which customers merely need ordinary outbound access? Which need stable source identity? Which need inbound reachability? Which need supplier recognition, reverse-DNS trust, mail reputation or public-sector audit evidence? Which can be moved to a lower-cost design without harming revenue? The answer often shows that a small minority of applications consumes a large share of the coexistence assurance budget. That minority should drive the product ladder, not be buried inside the average cost of access.

Wholesale contracts decide where ambiguity first lands

Wholesale contracts are written to turn complexity into saleable service. The buyer may purchase transit, access, backhaul, fixed-wireless capacity, enterprise handoff, route acceptance, static addressing, managed router support or emergency cooperation. The service description may say both address families are supported. The price usually arrives as a bundle. A bundle can be efficient, but it is also where the second stack often disappears.

The first hidden item is public identity. A wholesaler may include some IPv4 continuity, support IPv6 addressing and provide route evidence without pricing each input separately. "Included" then becomes a dangerous word. Public addresses have inventory, lease, transfer, reputation and opportunity costs. IPv6 support has device, monitoring and operational costs. Route evidence, reverse-DNS handling, contactability and emergency diagnostics require labour. If the retailer treats all of that as free, the first serious business customer turns an assumed feature into a dispute.

The second hidden item is fault boundary. A retail provider owns the customer conversation and often the CPE. The wholesaler controls upstream route acceptance, part of the public identity and sometimes the practical ability to diagnose where traffic failed. When an application breaks across address families, both sides can be partly right. The wholesaler can show availability; the retailer can show customer harm. If the contract defines packet delivery but not diagnostic cooperation, the retail support queue becomes the court of first instance.

The third hidden item is renewal leverage. A retailer that depends on a wholesaler's numbering, route evidence and emergency goodwill has less freedom to change supplier. BTW's earlier LACNIC leasing-contract analysis treated scarce address use as divided control between the party selling the service and the party holding the address position (btw.media). The same divided-control problem appears in dual-stack wholesale. The retailer sells continuity; the wholesaler may hold inputs without which continuity cannot be repaired quickly.

Wholesale renewal is the proper place to surface these costs. The buyer should ask whether the base charge covers IPv4 address continuity, IPv6-capable handoff, route-origin evidence, reverse-DNS and contact support, customer-facing diagnostic data, emergency routing cooperation, and evidence usable in enterprise SLA disputes. The seller should ask whether the retailer's device fleet, product language and support practices create avoidable upstream burden. Both should decide whether cost is recovered per active line, per business customer, per public identity, per managed application, per incident, or through a higher base charge.

None of this requires the wholesaler to itemise every packet. It requires the contract to stop pretending that address-family ambiguity is neutral. If the wholesaler can reduce the retailer's annual support and outage cost through better diagnostics, the wholesaler may deserve a premium. If the retailer's old CPE estate sends avoidable escalations upstream, the retailer should carry that cost. Bundled language does not abolish incidence. It merely delays the bargaining until after a customer is harmed.

The contract should also define what counts as proof during a fault. A wholesaler that says "traffic left our network" may be technically correct and commercially incomplete. A retailer that says "the customer was down" may be commercially correct and technically incomplete. Dual-stack service needs shared evidence: which family was preferred, which route was used, which public identity the counterparty saw, which CPE state applied, which security policy changed, and which customer application failed. Without an agreed evidence packet, incident calls become ritual blame exchanges. With one, the parties can assign elapsed harm to the controllable input that created it.

Retail bundles convert compatibility into tariff design

The retail tariff is where coexistence leaves engineering and enters household and business economics. A provider may preserve IPv4 through owned inventory, leases, static add-ons, shared translation or cloud features while expanding IPv6 through access equipment and upstream peering. The customer sees residential fibre, business internet, dedicated IP, managed security, hotel connectivity, public-sector service or a municipal bundle. The tariff decides who pays long before the customer reads an address plan.

In price-sensitive markets, the provider may be unable to raise the headline plan enough to recover coexistence costs. The burden then moves in quieter forms: slower CPE refresh, support rationing, paid static-address features, higher installation charges, less generous credits, delayed expansion or a wider gap between consumer and business tiers. The user may never hear the phrase dual-stack. The user experiences a service ladder.

Small businesses feel the ladder more sharply than households. A shop, clinic or lodging house may need more assurance than a consumer line but less than a full enterprise circuit. If the provider has no middle product, the customer is pushed down into ambiguous default service or up into a costly enterprise bundle. That misfit is itself a cost. It suppresses productive local services because the price of stable public identity and tested dual-stack behaviour is either hidden, over-bundled or unavailable.

The low-income market burden is related but not identical. BTW's LACNIC low-income analysis asks how fixed obligations divide by fragile receipts (btw.media). Dual-stack incidence asks which product should carry the burden. If coexistence is hidden in the base plan, all subscribers pay. If it is recovered through a business add-on, small firms pay. If it is absorbed in margin, future repairs and investment pay. If it is not recovered, service quality pays.

Honest tariff design does not mean turning protocol detail into a confusing menu. Most customers should not have to choose between address-family labels. They should choose assurance levels that match their economic use. Basic access should provide competent default reachability. A small-business plan should explain whether stable public identity, tested device behaviour and priority diagnostics are included. A revenue-critical application should carry an SLA that names address-family behaviour, public identity, monitoring, fault evidence and vendor cooperation. Wholesale resellers should know whether they are buying only capacity or also identity and recovery obligations.

The capital point matters. Lu Heng's argument that operators should stop apologising for IPv4 scarcity and treat scarce public identity as productive capital has a practical tariff implication (heng.lu). A provider ashamed to price public identity will give it away until scarcity forces rationing by delay, favour or frustration. A provider that treats it as capital can allocate it toward customers whose revenue justifies the assurance while letting lower-assurance uses benefit from IPv6 and competent defaults where appropriate.

The goal is not to make compatibility expensive for its own sake. It is to prevent a hidden cross-subsidy from undermining the network. Residential users should not unknowingly finance every business exception. Business customers should not discover after failure that the product they bought never included the identity they needed. The tariff should tell finance, support and customers what the bundle actually promises.

This is where the service bundle becomes a governance instrument without ever becoming a public policy. The provider can keep the retail offer simple while making the internal economics precise. A customer-facing label such as "business assurance" may hide technical complexity from the buyer, but it should not hide cost from the operator. Behind the label, the provider should know whether the price recovers a dedicated public IPv4 source, tested IPv6 paths, managed CPE, additional monitoring, vendor escalation rights, route evidence and shorter recovery obligations. If the bundle is cheaper than those inputs, the loss is not a marketing discount; it is an unrecorded transfer from future resilience to today's sales.

Devices make the second stack a depreciation problem

Customer premises equipment is where the abstract second stack becomes a depreciation schedule. The access network may support IPv6, but the installed device base may not support it reliably, visibly or uniformly. Some routers mishandle prefix changes. Some firmware exposes weak diagnostics. Some security defaults differ by family. Some older devices keep customers effectively IPv4-centred while newer replacements prefer IPv6 for selected destinations. Under one product name, support staff may face several service behaviours.

That split is expensive because equipment is not just hardware. It is procurement, inventory, installation labour, truck rolls, returned boxes, training, firmware management, help-desk scripts and customer tolerance. A fast refresh can lower long-term ambiguity but consume cash today. A slow refresh protects cash but pushes expected faults into operations. Either choice belongs in the annual coexistence unit. Capital pays upfront or support pays later.

The LACNIC region's variety makes this more than a technical preference. An urban fibre provider may amortise a device refresh across many subscribers. A rural fixed-wireless provider may treat each site visit as a material cost. An island provider may hold spares because delivery delay is part of outage risk. A public-sector service may require documented equipment behaviour. A hotel connectivity supplier may need devices that can support guest access, management interfaces, payment systems and back-office applications without creating inconsistent path selection.

Lu Heng's critique of the IPv6 escape narrative is helpful because it reminds operators that abundance in one address family does not abolish the cost of building an operating world around it (heng.lu). If the second world requires new devices, firmware policy, monitoring, training and support while the first world remains commercially necessary, the operator has not escaped scarcity. It has added a second depreciation track.

This is also why retail and wholesale contracts must name device responsibility. If the retailer owns CPE and sells the customer promise, it should carry the cost of predictable device refresh and accurate customer state. If the wholesaler supplies managed routers or depends on particular diagnostic data during faults, those obligations should be priced. If an enterprise customer chooses a cheaper unmanaged device despite revenue-critical needs, the SLA should not silently upgrade the provider's liability.

Device economics also exposes product mismatch. A cheap consumer router may be adequate for ordinary access and poor for a shop with cameras, payment terminals and remote support. A managed business router may look expensive until the provider prices fewer calls, clearer logs, policy parity and shorter restoration. A device that merely lists IPv6 support on a datasheet is not automatically cheaper than one whose behaviour is known across the full service life. The relevant cost is not purchase price. It is annual customer assurance.

Finance should therefore treat the CPE plan as a portfolio decision. Some devices can remain in service because their customers consume low-assurance access and create little dual-stack ambiguity. Some should be replaced early because they sit in businesses whose revenue depends on stable identity and rapid diagnosis. Some should be moved into a managed-device product where the customer pays for assurance directly. Some should be retired because their support cost now exceeds the remaining depreciation benefit. The technical inventory becomes a risk-weighted asset schedule. That is less elegant than a universal upgrade programme, but it is more likely to match the economics of a LACNIC-region provider with mixed customer income, uneven geography and hard limits on capital.

Vendor parity gaps turn coexistence into procurement hold-up

Dual-stack cost often hides inside vendor parity gaps. A router supports both families, but traffic-management features are richer on one. A firewall can filter IPv6, but policy presets, logs or threat feeds are less complete than the IPv4 process. A monitoring tool checks reachability without showing application fallback. A customer-management system has one "public IP" field even though the service now has several identity states. A cloud product offers IPv6 but charges separately for a public IPv4 source that a conservative counterparty still demands.

Each gap may look small in procurement. Together they become hold-up. Vendors gain leverage because coexistence expands the surface for licences, support tiers, consulting, upgrades, monitoring, managed firewalls and migration services. That does not make vendor spending illegitimate. Much of it is necessary. It does mean the buyer should treat a dual-stack strategy as a lifecycle cost, not as a feature checkbox.

The LACNIC-region provider often buys equipment, cloud services and software at global or hard-currency prices while selling connectivity into local tariffs. A licence gap priced in dollars can consume the margin from a small-business product group. A vendor support incident can turn a cheap device into an expensive one. A promised feature that remains incomplete for another year can force manual workarounds, extra support and customer exceptions. If finance does not allocate those costs to the product or customer that needs them, they land in general margin.

Lu Heng's account of why IPv6 was pushed is useful only if read as an incentive analysis rather than as a slogan (heng.lu). Complexity creates upgrade and advisory markets. Operators should therefore ask whether the vendor stack genuinely lowers the fully loaded annual cost of coexistence, or merely transfers expenditure from capital equipment into support, licences and fault response.

Procurement should test parity in operational terms. Are firewall logs equivalent across both families? Are support escalations equally mature? Are customer-facing diagnostics able to show path preference, fallback and public identity? Are security rules symmetric? Are route and DNS dependencies visible? Which features require extra licences? Which are promised but not production-stable? Which customer commitments would be breached if the weaker family failed? The vendor answer should be translated into money and assigned to a product, not left as a technical note.

One useful discipline is to price the workaround as if it were a product. If a missing vendor feature requires manual log correlation, a specialist support rota, a separate public-IP purchase, a temporary firewall rule or an exception register, that workaround has an annual cost and an owner. It should not be justified indefinitely by the phrase "until the vendor roadmap catches up". A roadmap is not a credit note. If the workaround protects a customer's revenue, it belongs in the customer's SLA or in the provider's premium bundle. If it protects only a vendor's weak parity, the procurement renewal should ask why the vendor is not bearing more of the cost.

This discipline can improve bargaining among wholesalers, retailers and enterprise buyers. A wholesaler that has invested in better dual-stack diagnostics may price that capability. A retailer choosing cheaper devices may accept more first-line support responsibility. An enterprise buyer demanding parity may pay for validated equipment and evidence. A public contract requiring both modernisation and legacy compatibility should fund both. The alternative is procurement theatre: a tender says "dual-stack", a datasheet says "supported", and the incident ledger later shows who actually paid.

Support queues reveal the costs invoices hide

The support queue is the most honest early-warning system for hidden incidence. Customers do not call to discuss address architecture. They report failed cameras, payment-terminal errors, remote-access trouble, inconsistent geolocation, blocked supplier portals, VPN failures, slow application starts, email reputation problems, or a service that works from one device and fails from another. Each call has a cost. Each unresolved call weakens trust.

Support cost is often pushed onto the weakest party in the chain. The wholesaler points to a clean circuit. The vendor asks for logs. The cloud platform shows a reachable service. The application supplier says its allowlist is unchanged. The retail provider still has the customer on the phone. The help desk becomes the absorber of incomplete contracts among upstreams, vendors, platforms and customer applications.

The provider can reduce that cost only by investing in visibility. Staff need tools that show customer device state, IPv4 public identity, IPv6 prefix state, recent configuration changes, route health, DNS answers, security policy hits and application symptoms without turning every call into a protocol tutorial. Scripts should ask business questions: is this a payment system, a camera, a supplier portal, a remote-work tool or ordinary browsing? The answer tells the provider whether the caller is buying convenience or revenue protection.

Support data should feed tariff and contract design. How many tickets involve IPv4-only counterparties? How many involve IPv6-capable devices with legacy applications? How many require vendor escalation? How many result in credits? How many are caused by product promises that were not priced? How many would disappear after a CPE refresh, better diagnostics or a different wholesale evidence obligation? These figures convert anecdote into incidence.

CGNAT belongs only in the background of this article. Shared translation is one way to stretch scarce IPv4 and can create support and attribution costs, but the hidden-tax treatment belongs elsewhere. The wider point is that even without dwelling on shared-address mechanics, dual-stack operation forces support teams to handle public identity, address-family selection, device capability, route evidence and application assumptions. The support queue prices ambiguity.

BTW's LACNIC customer-continuity analysis described network identity as relationship capital (btw.media). Support is where that capital is defended or wasted. A customer who receives a clear diagnosis, a suitable product choice and a short recovery path may accept a higher tariff. A customer who hears several suppliers blame one another will treat the network as unreliable even if the underlying infrastructure is sound.

The queue also protects the provider from false economy. A cheap wholesale deal that creates more escalations may cost more than a higher-priced deal with better route evidence. A cheap CPE fleet may raise annual support cost. A free static-address policy may consume specialist labour and scarce inventory. A premium dual-stack assurance product may look expensive until its lower support burden is measured. Support is not merely a complaint function. It is an accounting system.

The most valuable support metric is not total tickets. It is avoidable ambiguity per product. A high-ticket household plan may still be acceptable if calls are short, predictable and low-value. A small-business plan with fewer but longer dual-stack escalations may be underpriced because each case requires senior engineers, vendor contact and customer-credit negotiation. A public-sector or hotel service may create few incidents but carry high elapsed-harm exposure. Support reporting should therefore connect ticket type to revenue at risk, technical input, contract owner and prevention option. Once that connection exists, support stops being a cost centre pleading for more tools and becomes a source of pricing evidence.

Outage reconstruction prices elapsed customer harm

Normal service hides coexistence cost. Outages reveal it. The relevant measure is not only packet loss or technical availability. It is elapsed customer harm: the time from the first customer-impacting failure to the restoration of the recognisable service the customer bought. In a dual-stack environment that clock can lengthen because partial reachability disguises failure, fallback paths behave inconsistently, and each party can prove that part of its layer is alive.

Consider a hotel group. The public website may be reachable over IPv6. The payment processor may still rely on IPv4 allowlists. Guest Wi-Fi may use one path, back-office systems another, and cameras a vendor relay that behaves differently after a firmware change. The access provider can show the circuit up. The cloud dashboard can show green checks. The hotel still loses bookings or staff time. The recovery clock ends when bookings, payments and operations are usable again, not when one path responds.

LACNIC island and rural markets make elapsed harm especially visible. BTW's island-network dependency analysis framed the key question as whether the same public identity survives a changed physical path quickly enough (btw.media). The rural-connectivity scarcity article measured how fixed costs and repair time divide across sparse active lines and public-service anchors (btw.media). Dual-stack incidents combine those lessons. A partial failure consumes scarce support labour, emergency upstream time and customer patience while the provider discovers which identity failed for which application.

The annual coexistence unit should therefore include expected recovery cost: overtime, vendor support, temporary routing changes, emergency public-address features, customer credits, SLA penalties, support backlog, reputational damage, deferred installations and management time. Some items resist precise pricing. Ignoring them is worse. A provider that underprices high-assurance service will pay during the outage, often in the budget least prepared to absorb it.

Contracts should define recovery cooperation before the next incident. If the retailer depends on wholesaler route evidence, the wholesaler should supply timely diagnostic data. If the retailer owns CPE and customer promises, it should maintain accurate device and product information. If an enterprise SLA depends on cloud public identity or managed firewall behaviour, those vendor duties should be included. If a customer chooses a legacy application or a conservative supplier, the SLA should say whether the resulting compatibility cost is included or extra.

Lu Heng's registry-power-and-liability argument has a narrower operator-level analogue here: control over a critical input should be matched with some measured consequence for failure or delay (heng.lu). That does not mean unlimited liability. It means the party able to reduce elapsed customer harm should not be able to externalise the full cost onto the party closest to the complaint.

Outage drills can make the number visible. Select representative products: basic household access, a small-business plan, a hotel or clinic application, a public-sector service and a wholesale handoff. Simulate an IPv4 path issue, an IPv6 routing issue, a CPE firmware split, a cloud allowlist problem and a security-policy asymmetry. Measure functional restoration, not only network restoration. Then attach cost to elapsed time. The result may show that some products are too cheap, some wholesale terms too vague, some vendor contracts too weak and some customers under-insured for their own revenue risk. That discomfort is useful. It lets the renewal table reassign the charge before the next outage writes it by force.

The drill should also record which party could have shortened the clock. If the missing input was a wholesaler's route trace, the recovery term belongs in the wholesale agreement. If the delay was a customer vendor's quarterly change window, the customer should decide whether that risk is worth a higher managed-service product. If the bottleneck was a CPE model with weak diagnostics, the device plan should change. If a cloud public-IP feature was bought in panic at a premium, the architecture review should decide whether to pre-provision it or price it as an emergency service. Elapsed harm is not only a measure of failure. It is a map of bargaining power.

Registry discipline should lower recognition risk, not set tariffs

LACNIC's useful role in this economics is narrow. A number-resource registry can reduce uncertainty around records, proof of control, transfer history, contactability, reverse-DNS continuity, security assertions and route-adjacent evidence. Those functions matter because operators, wholesalers, lenders, enterprise buyers and counterparties need to know that a scarce public identity can be relied upon. Better recognition can lower route-acceptance friction, migration risk and buffers in wholesale or enterprise contracts.

The wrong role would be to turn coexistence into a tariff command. A registry should not decide whether a business customer deserves dedicated public IPv4, whether a local provider has modernised quickly enough, whether leasing or commercial use is morally attractive, or whether a retailer's device-refresh cycle is acceptable. Those are operator, customer, lender, court, contract and market questions. The registry's job is to make the common record reliable enough for those actors to make decisions without unnecessary uncertainty.

The distinction is central to Lu Heng's Bill of Rights of Uniqueness Coordination: the registry may record, coordinate and protect uniqueness; it may not rule (heng.lu). In dual-stack incidence, the economic translation is simple. Accurate records, proof of control, portable continuity and narrow dispute handling reduce the annual cost of coexistence. Broad discretionary language, unclear evidence expectations and mission creep add a registry-risk premium to a bill already paid through tariffs, support and capital budgets.

Running-Code Primacy gives the same discipline from the operating side (heng.lu). The number-resource layer exists because running networks need uniqueness, interoperability, proof, continuity and security-relevant metadata. It does not exist to supervise product pricing, business models, local customer mix or transition virtue. When a rule protects uniqueness and reliance, it can lower cost. When it turns operational change into permission theatre, it becomes part of the cost.

The design principle in Minimum Initial Specification, Localized Future Decision and Voluntary Adoption points in the same direction (heng.lu). Keep the common layer limited to deterministic, locally verifiable functions. Leave commercial evolution to the parties carrying the risk. LACNIC's region is too varied for a central institution to price the dual-stack burden across urban fibre, public-sector procurement, tourism systems, small firms, rural anchors, island recovery and enterprise cloud contracts.

This discipline does not make LACNIC unimportant. It makes the function more important and the discretion less defensible. A reliable ledger lowers the cost of proving identity. Stable reverse-DNS and security metadata can reduce migration friction. Transfer and leasing legibility can help product design. Dispute isolation can preserve customer continuity while a conflict is resolved. Each of those lowers recognition risk. None requires the registry to decide who should pay for a firewall licence, public-IP add-on, CPE refresh or support desk.

If the registry layer lowers uncertainty, the saving travels through wholesale renewals, enterprise SLAs, lender confidence and retail tariffs. If it raises uncertainty, the cost travels the same way. That is the registry's proper economic boundary.

NRS is useful where it improves holder-side bargaining

Number Resource Society belongs in this argument only in a proportionate way. It is not an access network, a replacement regional authority, a retail pricing board, a wholesale carrier, a device vendor or a public-address pool for every small business. Its future-facing value is that it organises holder-side vocabulary around rights, portability, exit, redundancy and accountability. In a dual-stack cost-incidence analysis, those concepts matter only if they lower avoidable uncertainty and improve the bargaining position of parties carrying coexistence costs.

NRS's public position frames decentralisation as systems engineering rather than institutional theatre (nrs.help). For a LACNIC-region retail provider negotiating with a wholesaler, device vendor, cloud platform or registry-adjacent counterparty, the practical value is not branding. It is a clearer outside option. A provider with portable proof, documented resource control and coordinated holder-rights language bargains differently from one that depends on a single opaque recognition path.

The NRS case archive also has incidence value because hidden costs survive by remaining isolated (nrs.help). One delayed correction, one uncertain route, one dispute over recognition, one transfer friction or one emergency continuity problem can be dismissed as local inconvenience. Patterns change bargaining. They allow operators, investors and enterprise buyers to ask whether registry-side or counterparty-side uncertainty should be priced explicitly into wholesale renewal, public-identity products or SLAs.

The danger is overstatement. If NRS were treated as a new central authority, it would reproduce the weakness it criticises. Its proper role is voluntary coordination, decentralised accountability and sharper holder-side bargaining. It can help small and medium networks resist cost shifting when product terms load all compatibility risk onto the retailer. It can make portability and exit less abstract. It can make continuity expectations easier to name. It should not decide every tariff or replace the local contracts through which service is actually delivered.

The continuity logic around LARUS One is relevant as a commercial analogy, not as a universal prescription. LARUS One separates public network identity from the delivery path (larus.net), and Lu Heng's note on network identity and customer continuity explains why renumbering becomes a business event once customers and counterparties rely on a stable public identity (heng.lu). For dual-stack incidence, the lesson is that identity continuity has measurable value. A provider that can preserve identity across supplier, cloud or path changes can reduce support shock and customer switching cost.

The NRS test is therefore concrete. Does holder coordination reduce the annual cost of proving control? Does it make portability more credible in wholesale contracts? Does it help investors classify registry-side risk? Does it let small providers demand clearer fault evidence and continuity terms from stronger counterparties? Does it support voluntary adoption rather than impose another mandatory scheme? If yes, it belongs in the cost-reduction column. If no, it is another meeting.

NRS should remain the only positive future-facing institution in this frame precisely because its best claim is not central command. It is the claim that the parties who bear number-resource risk need mechanisms, exit and accountability strong enough to negotiate with the institutions and suppliers around them.

The enterprise SLA makes cost allocation explicit

The most useful document after an incident is often not the engineering report. It is the enterprise SLA renewal. That is where the provider, customer, integrator and wholesaler can turn scattered costs into obligations. The customer has learned that "business internet" was too vague. The provider has learned that public identity, device behaviour, firewall parity and cloud egress were not separate details. The wholesaler has learned that route evidence and diagnostic cooperation may be part of the real service. The integrator has learned that application allowlists and vendor contracts can turn a protocol edge case into revenue harm.

The renewed SLA should start with service function, not protocol virtue. Which applications are revenue-critical? Which require stable public IPv4 source identity? Which can use IPv6 without counterparty change? Which need inbound reachability? Which require reverse-DNS, mail reputation, route-origin confidence or abuse-contact clarity? Which suppliers must be notified before identity changes? Which vendor controls the firewall policy, CPE firmware, application allowlist or cloud public-IP feature? These questions identify the economic surface that the generic product name concealed.

The next section should allocate recovery duties. The access provider can commit to customer diagnostics, device-state visibility and first-line triage. The wholesaler can commit to route-evidence response times and emergency cooperation. The integrator can commit to maintaining allowlists, vendor parity and application dependency records. The customer can commit to funding tested compatibility for legacy systems or accepting lower assurance where it chooses a cheaper tier. The cloud or managed-security supplier can be pulled into the evidence chain if its product is part of the service promise.

Price then follows obligation. Basic business access may include ordinary coexistence overhead. Dedicated public identity should be priced where the customer's application needs it. Managed dual-stack assurance should carry a higher tariff because it includes monitoring, diagnostics, fault evidence and recovery coordination. CPE refresh can be recovered through monthly equipment charges, installation fees or a premium service tier. Vendor parity gaps should be assigned to the party choosing the vendor or demanding the feature. SLA credits should be tied, where practical, to the party controlling the failed input.

This is where earlier BTW work on transfer-price transparency and route-object governance becomes relevant without turning the SLA into a registry debate. Price comparability helps when the provider must value scarce public identity (btw.media). Coherent route evidence helps when the customer needs assurance that a public identity can be accepted and trusted (btw.media). These are inputs to the SLA, not substitutes for commercial allocation.

The SLA will not make every allocation exact. Infrastructure contracts are incomplete. It can, however, stop the weakest party from becoming the default absorber of every unpriced dependency. If a legacy application forces IPv4 compatibility, the customer or integrator should decide whether the value justifies the annual cost. If IPv6 readiness lowers support cost for suitable services, the provider should capture part of the saving and share part through better pricing. If registry-side uncertainty raises recognition risk, the risk should be named rather than hidden in delays and buffers. If holder coordination improves outside options, that improvement should appear in stronger terms or lower risk premiums.

The enterprise negotiation is where dual-stack becomes measurable. It turns a support story into a contract map: which application needed which identity, which party controlled the relevant input, which evidence was missing, which product tier was underpriced, and which future payment will prevent the same bill from scattering again.

The investor review is where the bill is reassigned

The final scene should be a capital review rather than a protocol debate. The support engineer has found the incident. Finance has reconstructed the scattered cost. Sales has identified the customers most sensitive to public identity and recovery time. Procurement has listed device and vendor parity gaps. The wholesale team has prepared renewal options. The investor, lender or board committee now asks whether the provider is pricing coexistence or merely leaking margin.

The review pack should divide the annual coexistence burden into recoverable components. Public IPv4 inventory and leasing belong in a capital or product line. IPv6-capable device refresh belongs in depreciation and tariff design. Monitoring and security parity belong in assurance products. Support ambiguity belongs in training, tooling and product clarity. Emergency routing and vendor escalation belong in expected outage cost. Route evidence and registry recognition risk belong in wholesale and public-identity pricing. Customer credits belong in SLA design.

The committee should then compare products. Does basic residential access recover ordinary coexistence overhead without overburdening low-value uses? Does the small-business tier price stable public identity and tested device behaviour? Does the enterprise SLA recover diagnostics, vendor coordination and recovery duties? Does the wholesale agreement pay for route evidence and emergency cooperation? Does the CPE plan reduce annual support cost enough to justify acceleration? Does a cloud public-address feature belong in the customer price or in provider margin? The answers decide where incidence lands.

The investor should also ask what costs are being avoided by honesty. A provider that prices scarce public identity can preserve inventory for high-value uses. A provider that names device-refresh economics can reduce support surprise. A provider that defines SLA recovery cooperation can shorten outage harm. A provider that demands narrow registry discipline can lower recognition risk without pretending the registry is a tariff setter. A provider that uses holder-side coordination where it improves portability can bargain from a stronger position. Each improvement affects valuation because it protects cash flow and customer continuity.

The LACNIC lesson is specific. The economics of dual-stack cost incidence will not be decided by a declaration that one address family has won. It will be decided through wholesale renewals, retail bundles, device depreciation, vendor procurement, support queues, outage reconstruction, enterprise SLAs, tariff ladders and capital reviews. Geography matters because these channels differ across urban, rural, island, public-sector, tourism, enterprise and low-income markets. The method remains the same.

Measure the fully loaded annual coexistence cost for the customer, site or application whose revenue depends on both forms of reachability. Identify which contract creates the cost, which party can reduce it, which customer benefits from it, and which tariff or SLA recovers it. Keep the registry layer narrow enough to lower uncertainty rather than add rent. Use NRS-style holder coordination only where it strengthens voluntary bargaining, portability and exit. Then write the result into the renewal.

The dual-stack bill already exists. It is paid through invoices, margins, device cycles, support exhaustion, outage credits, customer churn and capital hesitation. The choice is whether it remains dispersed across budgets no one can defend, or whether the parties with the power to reduce it are made to see it, price it and carry it. In the LACNIC region, the decisive moment is not when a protocol is declared modern. It is when a contract finally states who pays for keeping both reachability systems alive.

Sources and further reading

These references provide the article's public doctrine and background context. They are used for institutional-economic framing, not for adopting any registry or official-sector narrative.