Summary

  • A bankruptcy court is equipped to supervise an estate sale: it can test the debtor's authority, give notice, hear objections, compare bids, approve a transaction, clear liens under applicable law and direct claims toward sale proceeds. Those are real powers over the debtor, the estate and parties before the court.
  • An RIR performs a different function. It authenticates the registered holder, applies published transfer conditions, prevents contradictory registration, updates the public record and changes access to services connected to that record. It does not distribute sale proceeds or decide the priority of secured creditors.
  • Nortel agreed in 2011 to sell 666,624 legacy IPv4 addresses to Microsoft for $7.5 million, or $11.25 per address. ARIN entered the transaction after the commercial bargain had been announced, and Microsoft agreed to place the transferred resources under a registration services agreement. The result joined a court-approved sale to an updated registry position without resolving every abstract argument about ownership.
  • Borders later marketed a legacy /16, received offers ranging from $1.50 to $12 per address and agreed to sell 65,536 addresses to Cerner for $786,432. Its proposed order was unusually explicit: the sale was subject to ARIN's consent and policies even while the court addressed value, liens, creditor notice and good-faith purchase.
  • A court order and a registry record are complementary evidence, not substitutes. The order can establish what the estate was authorized to convey and how claims are treated. The record determines who can maintain registration data, reverse DNS and, under the relevant agreement, routing-security credentials. Neither document alone proves global BGP reachability or clean reputation.
  • The workable boundary is procedural: registries should preserve a disputed status while litigation is live, accept final court evidence through a fast and reviewable path, verify exact resources and identity, then record the outcome. They should not reopen the bidding contest or reallocate creditor rights after the competent court has decided them.

The late actor controlled the last practical mile

The number that made the Nortel sale famous was $7.5 million. The fact that made it consequential was timing.

Nortel Networks was already in bankruptcy. Its advisers had identified a large inventory of legacy IPv4 addresses, found a buyer and put a signed agreement before the United States Bankruptcy Court for the District of Delaware. Microsoft had agreed to buy 666,624 addresses for $11.25 each. The court could see a debtor, a purchaser, specified blocks, consideration and a request to sell free and clear of claims. It had a familiar insolvency problem wrapped around an unfamiliar asset.

ARIN was not the auctioneer. It had not found Microsoft, valued the addresses, negotiated the price or ranked creditor claims. Contemporary accounts of the hearing record say the registry contacted the parties after the sale announcement. Yet ARIN controlled something neither the court nor the purchaser could treat as cosmetic: the recognized registration position and the services connected to it.

That is the apparent paradox. The actor that arrived late to the bargain could still affect whether the buyer received a clean operational handover. Microsoft could possess a court-approved contract, but a network team would also want accurate registration, current points of contact, reverse-DNS control and access to the relevant routing-security tools. Upstreams and counterparties would look for a public record consistent with the buyer's claim. The court could direct parties under its jurisdiction. It could not configure every independent network to accept Microsoft's routes.

The mistake is to turn this dependence into a contest over which institution is supreme. A court and an RIR are not rival title offices with identical mandates. They answer different questions. The court asks what the estate may sell, whether the process is fair to creditors and what happens to competing interests. The registry asks who is recognized in the number-resource record, whether the requested change is authentic and whether the resulting state remains unique and administratively coherent.

Nortel mattered because both questions appeared in one closing. Borders mattered because the second case wrote the boundary more explicitly. Together they offer a better rule than either institutional extreme: court orders must have practical effect, and registry changes must remain evidential rather than becoming a second insolvency trial.

What a bankruptcy court is actually built to decide

Insolvency concentrates claims. A company that cannot pay everyone may have secured lenders, trade creditors, employees, landlords, tax authorities and contractual counterparties asserting rights against a limited estate. Selling an asset outside the ordinary course is therefore not simply a deal between seller and buyer. The sale can alter what remains for creditors and whether liens follow the asset or attach to proceeds.

That is why a court-supervised sale uses notice, objections, evidence and an order. Under section 363 of the United States Bankruptcy Code, a debtor in possession or trustee may seek authority to sell property of the estate. The court can assess the business justification for the transaction, the adequacy of the marketing, whether a higher or better offer exists, the treatment of liens and whether the purchaser acted in good faith. It can retain jurisdiction over disputes arising from the sale and direct that eligible interests attach to the cash proceeds in their established priority.

These powers fit an IPv4 transaction surprisingly well. Scarce address space can carry substantial value. A hurried private disposition to an insider could deprive creditors of that value. A creditor might assert a security interest in the debtor's intangibles. A purchaser may refuse to pay full value if old claims can follow the resource. The court supplies a forum in which those objections can be heard together rather than pursued against the buyer one at a time.

The court is also positioned to test corporate authority. It can determine whether the debtor entity before it controls the claimed interest, whether the board or restructuring officers authorized the agreement and whether the signatory may bind the estate. If several affiliates used a block, it can examine which debtor is selling and whether another estate entities. If the transaction is an auction, it can approve bidding protections and evaluate the sale process. If it is a negotiated sale after broad marketing, it can decide whether the proposed bargain still represents the highest or otherwise best value.

None of this requires the judge to operate a routing registry. The court does not need to decide which ASN should originate each prefix, what a ROA's maximum length should be, which nameservers should receive an in-addr.arpa delegation or whether a block has a poor email reputation. Those are not creditor-priority questions. They become conditions of delivery, valuation and post-closing performance.

The legal order is therefore powerful but scoped. It can make the sale binding on the estate and on interests reached by the order. It cannot make the rest of the Internet observe a new BGP route. Nor can it guarantee that every registration service will change automatically. Good transaction design acknowledges that limit before the hearing rather than discovering it after money is committed.

Nortel turned an address inventory into a bankruptcy asset

Nortel's United States debtors entered Chapter 11 in January 2009. By March 2011, Nortel Networks Inc. sought approval to sell what its motion described as approximately 666,624 legacy IPv4 Internet numbers. The sale motion attached the purchase agreement with Microsoft and asked the Delaware bankruptcy court to authorize a $7.5 million transaction.

The motion's commercial logic was plain. IPv4 supply was limited, the addresses could attract value, and an orderly sale could return money to the estate. The agreement specified the resources rather than purporting to sell an undefined entitlement to whatever addresses appeared under Nortel's name. It allocated risk through familiar sale language, including an as-is framework and provisions addressing claims and assumed liabilities.

At $11.25 per address, the transaction did more than produce cash. It supplied a public price for a scarce resource that had long been discussed as if it were merely issued administrative space. A court was being asked to approve the transfer of an economically valuable bundle of rights. That did not answer every property-law question, but it made denial of the market fact impossible.

The careful description is important. The court approved the transaction presented to it. It did not issue a universal judgment that every IPv4 address is absolute property against every person in every jurisdiction. Later commentary sometimes inflates the order into that proposition; other commentary minimizes it as if nothing was sold. Both readings exceed the record. Nortel conveyed defined interests in specified legacy numbers, and Microsoft paid for them. The final arrangement also recognized the registry relationship.

The distinction between an interest and an absolute entity was not wordplay. IPv4 value depends on overlapping systems. The seller may have an exclusive claim against private competitors to use and convey its position. ARIN may maintain the authoritative regional registration and condition certain services on an agreement. Network operators independently choose routes. Courts enforce contracts and orders within jurisdiction. The valuable thing is a coordinated bundle, not a metal token sitting in a vault.

Nortel's sale brought that bundle into view because the insolvency proceeding forced every entity to state what it needed. Creditors needed value and claim finality. Microsoft needed delivery and operational continuity. ARIN needed the resulting record to fit its policy and contractual framework. The court needed an order capable of closing the sale without pretending to command the Internet.

ARIN arrived after the bargain but before clean completion

ARIN's own archived April 2011 announcement says Microsoft received court approval for a transfer in which the resources would be placed under a registration services agreement between ARIN and Microsoft. That sentence captures the settlement of a potentially larger conflict.

Nortel had legacy space dating from before ARIN's formation and did not approach the sale as an ordinary holder under a current registration contract. Microsoft nevertheless wanted the registry relationship that would make the new position legible and manageable. ARIN wanted transfers in its service region to follow a recognized policy path. The parties reached an arrangement under which Microsoft entered a Legacy Registration Services Agreement.

This was not ARIN ratifying the sale price. Nothing in the public record suggests that the registry was asked to decide whether $7.5 million was adequate for Nortel's creditors. It was not a secured lender, auction consultant or valuation expert. Its interest lay in registration, transfer conditions and the agreement governing future services.

Nor was the court simply ordering ARIN to surrender every policy position. Accounts of the April 26 hearing emphasize that the parties avoided litigating the full scope of ARIN's authority. The court approved the amended bargain. Microsoft accepted a registry agreement. The resulting order preserved the transaction without generating a sweeping answer to questions that did not need to be decided for closing.

That restraint is the enduring lesson. Commercial law often advances through settlements narrower than the arguments surrounding them. Nortel did not establish a complete law of Internet numbers. It demonstrated that a bankruptcy estate could realize value from a legacy IPv4 position and that an RIR record could be aligned with the court-approved transfer through agreement rather than institutional veto.

The phrase "late arrival" should not imply that ARIN's work was optional. It identifies a sequencing defect. Registration conditions should have been mapped before the price became news. If the buyer's agreement, eligibility or operational access had proved impossible, the estate could have lost time and value. The registry's necessary role belonged in diligence and closing design, not as an improvised dispute after selection of the buyer.

Borders showed the same boundary in the text of the order

The Borders case arrived later in 2011 and is even more useful because the documents state the division of functions in detail. Borders had closed its stores and was liquidating. A legacy /16, 170.71.0.0/16, had not been included in an earlier intellectual-property sale. The debtor's agent, Streambank, prepared an offering memorandum and contacted Internet service providers, network companies, registries and technology firms.

According to the Borders sale motion and proposed order, the marketing reached thousands of potential buyers. Offers ranged from $1.50 per address to Cerner Corporation's $12. Cerner agreed to pay $786,432 for all 65,536 addresses and to place a ten-percent deposit. Borders presented that offer as the highest and best result.

The sale was not identical to the earlier auction of Borders' trademarks and other intellectual property. The address block went through a separate marketing process followed by a court approval request. That factual difference matters. "Bankruptcy auction" is often used loosely for any court-supervised sale, but courts may approve formal auctions, stalking-horse processes or negotiated transactions after market testing. The accountability question is whether notice, exposure to competition and judicial review gave the estate a defensible result.

The proposed Borders order addressed classic insolvency matters. It found fair and adequate consideration, arm's-length negotiation and good faith. It authorized transfer of the debtors' interests free and clear of covered liens and claims, directed those interests to proceeds and protected Cerner from specified successor liability. It retained jurisdiction to enforce the sale and resolve related disputes.

Then the order drew a line. Its references to transfer meant transfer of the debtors' interests in the addresses. More importantly, paragraph 14 made the transaction conditional on ARIN's consent and stated that the debtor and purchaser had to comply with ARIN's published transfer policies before the transfer could be effectuated. The same document that used strong sale language also refused to treat the registry as irrelevant.

Borders therefore rebuts two simplistic positions at once. The IPv4 interest was valuable enough to market separately, compare bids and sell for more than three-quarters of a million dollars. Yet a sale order described as conveying good and marketable title still contained an explicit ARIN condition. The court could clear creditor claims; operational finality required the registry transition.

The court order answers questions the registry cannot

Suppose a lender says its security interest covers all of a debtor's intangible assets, including IPv4 rights. An RIR is not the proper forum to decide the lender's priority against the estate. Its staff do not administer the Bankruptcy Code, conduct a claims process or distribute proceeds. A registration hold may preserve the status quo while the dispute is presented elsewhere, but it is not a judgment on secured-creditor law.

Suppose a disappointed bidder says the trustee accepted a lower offer from an insider. Again, the registry cannot repair the sale process by selecting a different buyer. The bankruptcy court can hear evidence about marketing, relationships, consideration and good faith. If the court rejects the sale, there may be no registrable transaction. If it approves a final sale after objections, the registry should not quietly rerun the auction at the service desk.

Suppose several debtor affiliates claim the same block. The court can identify which estate holds the interest, construe intercompany agreements and authorize the correct entity to act. Registry history may be decisive evidence, but the public record may show a long-defunct name or an affiliate that operated the network without holding the economic interest. The registry can explain what its record says; the court can decide the legal contest among parties.

The order also creates a remedy structure. Covered liens may attach to proceeds. The purchaser may receive protection against claims that would otherwise follow the asset. The court can enforce delivery obligations, interpret the agreement and address a seller that refuses to cooperate. An RIR's transfer policy does not provide a substitute creditor distribution system.

These functions justify deference to a final order on the matters it decides. Deference does not mean blind data entry. The registry must still authenticate the order, match the named entities to the correct accounts and confirm the exact prefixes. It should also determine whether the order is final, stayed or subject to a conflicting injunction. But once those checks succeed, the registry's task is implementation, not a collateral appeal.

The registry record answers questions the court should not

The registry has the best view of its current administrative state. It knows the organization identifier associated with a block, the authenticated points of contact, the applicable agreement, pending transfer requests and whether another claimant has raised a resource-status dispute. It can coordinate an inter-RIR change and prevent the same prefix from being recorded in incompatible positions.

ARIN's current Number Resource Policy Manual requires the source in an in-region specified-recipient transfer to be the current registered holder and not to be involved in a dispute over the resource's status. Its transfer guide accepts finalized court orders among the evidence relevant to mergers, acquisitions and reorganizations and requires a chain of registration when a current registrant no longer exists. These are record-integrity questions with a direct connection to a reliable handover.

The registry also controls or coordinates services that become operationally material. ARIN's current Registration Services Agreement describes an exclusive right to be the registrant within its database, a right to use the included resources within that database and a right to transfer registration under policy. The agreement assigns authority over points of contact and directory data. It also places the resources within a service framework that includes registration management.

The record is not a route command. BGP speakers do not ask a bankruptcy judge or ARIN for permission before processing every announcement. Yet operators, transit providers and security teams use registration evidence when validating letters of authorization, responding to abuse, constructing filters and resolving incidents. A mismatch between legal buyer and registered holder can delay or defeat deployment even if packets are technically capable of flowing.

The right approach is to give the registry strong authority over record coherence and weak authority over unrelated commercial merits. It should reject forged requests, ambiguous prefix lists and contradictory final orders. It should not decide that a court-approved buyer paid too little, that creditors deserve a different allocation or that the debtor should have preferred another bidder. Those judgments are outside the registration function.

Operational usability is a stack, not a judicial sentence

An IPv4 block becomes useful through several aligned surfaces. Registration is one. Routing is another. Reverse DNS, routing-security credentials, Internet Routing Registry entities, geolocation, abuse contacts and reputation systems add further layers. The buyer must know which of these will move automatically, which must be rebuilt and which sit outside both court and RIR control.

ARIN's transfer best-practices guidance illustrates the point. Before a transfer, a source should remove transferred prefixes from relevant ROAs, review route objects and coordinate reverse DNS. On completion, the recipient must create its own ROAs and authoritative routing records and verify reverse-DNS delegations. The source's RPKI certificate is reissued to reflect the change, but the recipient is responsible for the new authorizations.

A bankruptcy order cannot write those ROAs. It can require the seller to cooperate or authorize a representative to execute necessary documents. The registry can make the recipient eligible to create credentials. The buyer's engineers must still state the intended origin AS and maximum prefix length. Relying networks then apply their own route-origin validation policy.

Reverse DNS has the same division. The court can sell the debtor's interest and order delivery. ARIN can change who manages the delegation. The recipient or its DNS provider must operate correct nameservers and PTR records. Cached data and TTLs may delay visible change. A sentence declaring the sale effective does not answer a DNS query.

Reputation is even further removed. A clean sale does not erase prior spam, malware or abusive hosting associated with the addresses. Blocklist operators and large platforms maintain their own evidence and removal rules. Geolocation providers may continue to place the range at the seller's facilities. Upstreams may retain stale filters. These external dependencies affect value, but they are not liens that necessarily disappear under a sale order.

This is why the phrase "free and clear" can mislead technical buyers. It is potent legal language concerning specified interests reached by the order. It is not a warranty that every route will propagate, every mailbox provider will accept traffic or every third-party dataset will update. Transaction documents should separate legal claims from operational residue.

A record change should not become a second trial

The registry's strongest argument for independent review is fraud prevention. A document bearing a docket number may be incomplete, superseded or presented by someone without authority. Prefixes can be omitted or substituted. A sale order may approve an agreement subject to unsatisfied conditions. An appeal may be stayed. The named purchaser may use a subsidiary not shown in the registry request.

Verification is therefore justified. Staff should obtain the filed order from a reliable court source or authenticated counsel, inspect the schedule of resources, confirm effectiveness, match the seller and buyer, identify conditions and preserve an audit trail. If language is ambiguous, the parties can seek clarification from the court. None of this reopens the merits.

Trouble begins when record verification turns into a new adjudication. A registry should not require a trustee to prove again that the selected price maximized the estate. It should not ask creditors to relitigate liens through private correspondence. It should not impose a new theory of which affiliate economically deserves the block after the court has resolved the issue. Nor should staff use a delay of indefinite length as leverage to change commercial terms.

The distinction can be expressed as a counterfactual. Ask whether the registry's concern would still matter if the addresses had no market price. Exact prefix identity, authentication, duplicate registration and conflicting orders would still matter. Bid adequacy, creditor priority and purchase price would not. The first group protects a reliable registry. The second belongs to the legal proceeding.

Where current policy imposes recipient qualifications, the registry may apply those published conditions, as the Borders documents expressly anticipated. But those conditions should be objective, known before bidding and capable of pre-clearance. A buyer should not discover after the auction that an unwritten view of its business model prevents recognition. Predictability protects both registry integrity and creditor value.

Continuity risk appears in the gap between closing and recognition

The most dangerous transaction state is not necessarily rejection. It is ambiguity after a legal closing but before an operationally complete handover.

If purchase money has moved and the court order is effective while the old organization remains registered, the parties face competing signals. The buyer may believe it owns the transferred interest. The seller's former contacts may retain account access. Upstreams may see a new origin but old registration. Abuse reports go to a company being dissolved. ROAs may validate the seller's ASN or disappear before the buyer creates replacements.

Bankruptcy makes the gap more severe because the seller is vanishing. Employees leave. Credentials are disabled. Vendors terminate service. The estate's professionals may understand the sale but not the network. If a required signature or account recovery is postponed, there may be no ordinary operating team left to provide it later.

The purchase agreement should therefore treat registry completion as a designed closing condition or a simultaneous deliverable, not a clerical follow-up. Funds can remain in escrow until specified registration evidence appears. A portion can be held back for transition duties. The order can authorize named representatives to complete the transfer even if normal corporate officers depart. The prefix schedule should be final before the hearing.

Continuity planning must also allow staged routing. The buyer may need the old origin to remain authorized briefly while new filters and credentials propagate. That does not mean the seller retains the economic bargain. It means network state has latency. A controlled make-before-break plan can preserve reachability while registration and routing authorization converge.

The registry should publish or privately provide decisive timestamps: request accepted, dispute hold applied, court evidence verified, recipient conditions satisfied, record changed and credentials reissued. The buyer and estate can then distinguish registry delay from their own missing documents. A single status reading "under review" is inadequate when a live network or expiring estate depends on the result.

"Ownership" is too blunt for the closing checklist

Debates about whether IPv4 addresses are property often produce more heat than delivery precision. A buyer does not close on a philosophical noun. It closes on defined rights, evidence and operational capabilities.

The contract should say exactly what the seller conveys: its rights and interests in listed prefixes; its recognized registration position to the extent transferable; authority to request a registry change; associated records or data that can lawfully be delivered; and transition cooperation. It should identify what is excluded, such as customer data, domain names, equipment or claims against third parties.

The court order should address estate authority, creditor interests, good-faith purchase, effective time and the binding effect on parties reached by the proceeding. If the sale remains subject to registry conditions, that should be explicit. If the order directs a representative to act, the grant should be clear enough for authentication.

The registry receipt should identify the exact prefixes, source, recipient, applicable agreement and completion time. It should not call itself proof of price or creditor discharge. The buyer's operational receipt should separately record new contacts, reverse-DNS control, ROAs, route objects and stable BGP observations.

This vocabulary avoids a false binary. One need not decide that the court owns the Internet or that the registry owns every address. The estate can convey a valuable, enforceable interest. The registry can maintain the recognized record. Operators can decide routes. Each statement is meaningful without swallowing the others.

Nortel and Borders are strongest when read this way. They are not magic precedents that eliminate RIRs. Nor are they mere paperwork that leaves the estate with nothing of value. They show a divisible architecture in which legal transfer and registry recognition are distinct but must be synchronized.

The Borders order was a warning against institutional overclaim

The Borders proposed order contains language that appears to pull in opposite directions. It authorized a sale of the debtor's interests free and clear, described the purchaser's rights in strong terms and barred covered claimants from interference. It also conditioned effectiveness on ARIN's consent and compliance with ARIN policies.

That combination was not incoherent. It reflected two sources of risk. Creditors and other claimants could attack the asset from the legal side. ARIN could decline to change the registration from the administrative side. Cerner needed protection against both before paying the full price.

The danger is that each institution may use the other's recognition to enlarge its own claim. A registry might say that because the court order included ARIN conditions, the court accepted unlimited registry authority. The text does not support that leap; the conditions avoided deciding a larger dispute. A purchaser might say that because the court conveyed the debtor's interests, ARIN had no remaining record-integrity role. Paragraph 14 says the opposite.

The restrained reading is also economically superior. If a registry can override a final sale on any ground it chooses, insolvency buyers will discount or avoid IPv4 assets. Creditors lose value. If a court can order registration without regard to exact identity, duplicate claims or technical handover, the public record loses reliability. Operators bear the resulting risk.

Institutional boundaries preserve value because they make diligence possible. A bidder can assess court risks separately from registry risks. It can obtain pre-clearance on objective eligibility, review the docket for liens and objections, and price operational cleanup. Unbounded authority cannot be diligenced because every question collapses into discretion.

The Borders record should therefore be read as a design document for coordination. It says, in effect: the court will decide this sale and clear covered claims; the registry's published conditions still matter; neither side needs to deny the other for the transaction to close.

A fast protocol for court-directed IPv4 transfers

Future insolvencies should not improvise the Nortel sequence. Courts, trustees, buyers and RIRs can use a standard protocol without turning the registry into a creditor tribunal.

First, the estate should identify number resources at the beginning of asset review. It should preserve account access, points of contact, agreements, historical registration evidence, reverse-DNS data, ROAs and route objects. Waiting until employee termination destroys evidence and leverage.

Second, the trustee should notify the responsible RIR before marketing closes. Notice is not a request for valuation approval. It asks the registry to confirm the current record, applicable transfer path, agreement status, pending disputes, exact evidence requirements and realistic service times. The response should be written and stable for a defined bidding period.

Third, the registry should apply a narrow preservation flag if litigation or competing claims exist. The flag should block unauthorized change without erasing public data, revoking operational services or prejudging ownership. Its reason, scope, reviewer and expiry should be recorded.

Fourth, bidders should be allowed objective pre-clearance. Identity, recipient account readiness, regional path, agreement form and any published qualification can be tested before the auction. Pre-clearance should be portable to the listed prefixes and should not expose the bidder's maximum price.

Fifth, the sale motion should attach an exact CIDR schedule and describe registry conditions accurately. It should distinguish the debtor's interest from a warranty of universal routability. Notice should reach the RIR, known secured parties and any identified adverse claimant.

Sixth, the final order should state the effective time, treatment of interests, authority to sign, conditions precedent and whether a stay applies. It should direct cooperation but avoid technical commands that cannot be executed as written.

Seventh, the RIR should verify and record the final outcome within a published time. Any refusal should identify the precise unmet condition and a prompt review path. It should not rely on an unexplained conclusion that the sale is inconsistent with community expectations.

Finally, the parties should complete an operational acceptance test: registration, contacts, reverse DNS, RPKI, IRR, routing visibility, filtering and abuse channels. Only then can the buyer say that the block is usable, rather than merely purchased.

The registry should record finality, not manufacture it

The proper role of an RIR in a bankruptcy sale is neither passive copying nor sovereign approval. It is reliable recognition.

Reliable recognition begins before the order. The registry tells the estate what its record shows and what objective conditions apply. During a live dispute, it preserves the existing state and avoids irreversible action. After a final order, it authenticates the document, matches resources and parties, confirms conditions and updates the record. If two orders conflict, it asks the competent court for clarity rather than choosing a commercial winner privately.

This role is essential precisely because the registry does not create the underlying legal finality. Its credibility comes from accurately reflecting authoritative events within its scope. A land registrar does not conduct every sale; a securities depository does not decide every corporate dispute; a vehicle registry does not distribute bankruptcy proceeds. Each becomes valuable by recording valid outcomes consistently.

IPv4 registration is not identical to those systems, but the institutional principle travels. Recordkeepers become dangerous when dependence on their records is converted into a claim to decide every underlying right. They also become useless if they ignore legal reality and preserve a dead company's name forever.

Number Resource Society can advocate a cleaner division of responsibility: portable records, deterministic state transitions, authenticated orders, visible holds, bounded review and a separation between record service and commercial judgment. It can support affected members and publish evidence-led analysis of where that boundary fails, but it does not authenticate court orders, place holds, update registrations or operate a continuity service. Courts remain the institutions that settle legal claims and bind parties; the responsible RIR verifies and records the resulting registration change.

The improvement is to prevent a centralized service provider from turning implementation into a second, unbounded trial, not to substitute NRS for either institution.

In that model, a court order enters the record as evidence with a defined effect. The registry verifies rather than philosophizes. The buyer can prove the chain from estate authority to recorded control. Operators can inspect current status without being told that registration guarantees routing. Error correction remains possible through a documented process.

The result is thinner governance and stronger continuity. Fewer institutions claim total authority. More evidence survives closing. Creditors receive clearer value. Buyers know what they are purchasing. Networks receive a registration state they can trust.

What the two sales finally proved

Nortel proved that IPv4 scarcity could surface as realizable bankruptcy value. A large legacy inventory was specified, marketed and sold to Microsoft for a disclosed price. The court approved the bargain. ARIN's late participation did not erase the sale, but it changed the form of completion by bringing the buyer into a registration agreement.

Borders proved that the boundary could be written directly into sale documents. The court handled marketing evidence, consideration, claims, good faith and transfer of the estate's interests. The order also made ARIN consent and policy compliance a condition. Cerner was not asked to choose between legal title and registry recognition; it needed both.

Neither case proved that an RIR owns IPv4 space. Neither proved that a court can guarantee routability. Neither established that every legacy holder has the same contract or that every jurisdiction will characterize the interest identically. Those limits do not make the cases weak. They make them useful.

The practical hierarchy is now clear. Courts decide the estate, the sale and the claims properly before them. Registries decide whether and how their recognized record changes under a bounded, published service. Network operators decide what routes they accept. Buyers must assemble all three forms of finality.

When those layers are confused, the late-arriving registry appears either all-powerful or irrelevant. It is neither. It controls the last practical mile of an administrative handover, and that mile can determine whether a court-approved asset is ready to operate. The answer is not to let the registry retry the case. It is to bring record requirements into the transaction early, then require the recordkeeper to implement legal finality accurately and on time.

Five failure tests expose the boundary

The division becomes clearest when tested against failure rather than institutional description.

First, imagine that a creditor entities before the sale hearing and produces a perfected security agreement covering the debtor's intangible assets. The registry should preserve the record while the objection is resolved. It should not decide the creditor's priority. The court can determine whether the sale may proceed free and clear and whether the creditor's interest attaches to proceeds. Once a final order identifies the treatment, the registry can verify the resource schedule and implement the recognized change.

Second, imagine that no creditor entities, but the CIDR schedule in the order includes a /17 that the debtor never held. Here the registry should refuse that portion. The court's authority over the estate does not enlarge the estate's registration position by typographical accident. The parties should return for a corrected schedule or show the missing chain. Record integrity is not disrespect for the order; it is what gives the valid portion meaning.

Third, imagine that the order is correct but the purchaser fails a published recipient condition disclosed before bidding. The registry may apply that condition. The commercial documents should then determine whether closing fails, an alternate eligible affiliate can receive the registration, or the estate returns to the market. What the registry may not do is invent a new condition after selection of the buyer and call it a technical detail.

Fourth, imagine that registration changes cleanly but a former customer continues announcing a more-specific route. The court order and RIR record may both be effective while traffic still follows the customer because longest-prefix routing favors the more-specific announcement. The remedy lies in transition obligations, upstream filters, authorization withdrawal and, if necessary, enforcement against the customer. Reversing the registry record would not by itself repair the route.

Fifth, imagine that the buyer becomes registered and creates correct ROAs, but a mail provider continues blocking the range because of the debtor's history. No institutional contradiction exists. The sale can be legally final, the record accurate and the route valid while reputation remains poor. The buyer's recourse depends on warranties, holdbacks and the provider's remediation process, not on asking the bankruptcy judge to declare the addresses reputable.

These tests show why one certificate of completion is impossible. Each failure belongs to an actor with relevant evidence and a remedy suited to the problem. The market becomes safer when the court, registry, operator and buyer can each say exactly what has and has not been completed.

Sources