Summary

  • The existing RIR transfer logs are valuable custody records. Their common JSON publication design can identify transferred IPv4, IPv6 or AS Number sets, source and recipient organisations, registry pair, transfer date and broad transaction type. They can support provenance checks and research into post-transfer routing. They are not a complete market record.
  • A completed row omits the contract price, currency, fees, contract date, request date, time spent with each registry, rejected and withdrawn cases, pending age, reason for failure, broker coverage and transactions that changed operational control without a recognised transfer. Successful transfers are therefore a selected numerator with no stable denominator.
  • Price confidentiality does not require price blindness. Registries, brokers and escrow providers could report quarterly medians, quartiles, volume-weighted means, block-size bands, interquartile ranges and coverage rates only after a minimum cell count is reached. Publication should be delayed and separated from named transaction rows so an aggregate cannot be reverse-engineered into a party's contract.
  • Delay must be split by stage and controller. Calendar time from complete request to record change, registry-controlled days, party-response days, bilateral handoff time, fee and contract time, and age of pending cases answer different questions. A single average completion time allows one institution to attribute the tail to everyone else.
  • Rejection reporting needs an explicit case unit and a reconciliation: opening pending cases plus new complete requests should equal approvals, rejections, withdrawals and closing pending cases after accounting for reopened or merged cases. Standard reason categories and appeal outcomes can be public without exposing applicant documents.
  • Number Resource Society can provide the missing public-interest layer as a minimal ledger and independent aggregation service. It can publish definitions, collect signed but confidential price observations, reconcile RIR performance, preserve revision history and make records portable while refusing to broker deals, set prices or turn its own certificate into a new gate.

The row for the winner is an attractive illusion

A transfer log creates the satisfying appearance of a market seen whole. There is a date, a block, a source, a recipient and perhaps two registries. The resource moved. The public record changed. An analyst can count rows, sum addresses and draw a line through time.

Everything difficult happened outside the row.

The seller may have marketed the block for months. Several buyers may have withdrawn after learning that their registry path was unavailable. The eventual buyer may have obtained pre-approval, renegotiated price when reputation checks found abuse history, waited for corporate documents, paid a broker and escrow agent, and then watched the two registries coordinate. A rejected applicant leaves no completed row. A buyer that could not satisfy a needs test may never file. A seller that leased operational control without changing the recognised holder may never appear.

The final transfer date is therefore a publication event in a selected population. It proves that one record changed. It does not prove that the process was quick, open, competitive or consistently administered. It does not reveal whether the same block sold for USD 10 or USD 50 per address, whether a small buyer paid a premium, or whether a regional incompatibility reduced the bidder pool.

That distinction matters because a registry is not only observing the market. Its decision is part of completion. The buyer and seller can agree on price, yet closing often depends on the authoritative record being updated. A log of approvals is partly a log of the institution's own exercise of power. Members cannot evaluate that power if they see only cases that survived it.

The absence is most serious in thin markets. Ten successful transfers may look like steady activity. If they came from eleven complete requests, the desk appears predictable. If they came from one hundred requests, with long delays and unexplained withdrawals, the same ten rows describe a restrictive system. If brokers screened hundreds more because a route was known to be closed, even the registry denominator is incomplete.

This does not make the completed log deceptive. It makes its purpose narrower than its visual authority. A custody log answers: which recognised record changed, from whom to whom, for what resource, and when? A market transparency statement answers: who could participate, what they paid, how long cases took, which cases failed, why they failed, and how much of the market the data cover?

Both are needed. Combining them in one public row would expose contracts and create security risks. Treating the first as a substitute for the second protects the institution from scrutiny. The correct design keeps transaction identity and market accountability in separate, reconcilable publications.

What the common RIR logs actually record

The five registries have done important technical work toward comparable publication. ARIN's description of the NRO transfer-log format says the common JSON design covers IP address and AS Number transfers within and between RIRs. Each registry publishes a cumulative transfers_latest.json file and dated archives. That is a stronger foundation than isolated web tables because researchers can retain versions and inspect change over time.

The live feeds are available from ARIN, APNIC, RIPE NCC, LACNIC and AFRINIC. A field review on 14 July 2026 found a recognisable shared core: original and transferred resource sets, source and recipient organisations, source and recipient RIRs, transfer date and transaction type. The resource set may contain IPv4 networks, IPv6 networks or AS Numbers. Organisation entries can include name and country code.

The feeds are not identical. ARIN and LACNIC include a source-registration date on some records; the other current feeds do not expose that field. Current files use broad labels such as resource transfer and merger or acquisition. The latest transfer dates visible in the files also differ substantially, so a consumer cannot assume that downloading five files at one moment produces five equally current populations.

The common core is well suited to record provenance. A network operator can ask whether a block moved after a historic allocation. A researcher can measure address volume by destination region, compare block sizes or examine routing after the transfer date. The 2017 study On IPv4 Transfer Markets combined published RIR transfers with routing and other data to study whether transferred space became used and to infer unreported movement. That is evidence of the logs' research value.

But the common fields describe the completed resource state, not the transaction's economic or administrative path. They contain no contract price, currency, broker, escrow date, request timestamp, completeness date, approval timestamp, rejection state, withdrawal state, pending state, review reason or appeal. A merger and an arm's-length sale can both move a resource while reflecting different markets. A resource-transfer label does not prove that money changed hands.

Even the date needs care. It is the recorded transfer date, not necessarily the contract signature, price agreement, payment, routing change or first use. Treating it as the sale date can move an observation into the wrong market period. Treating every transferred address as sold can convert corporate reorganisations or gifts into invented turnover.

The present logs are thus a public skeleton: unusually useful, cumulative and machine-readable, but built for custody. Transparent market analysis must add a second layer whose definitions do not corrupt that purpose.

The unit of observation is not yet stable

Before adding fields, the registries need to say what one case is. This is more difficult than it sounds.

A commercial agreement can contain several prefixes. A prefix can be divided into several transfer records. One corporate transaction can move IPv4, IPv6 and AS Numbers. ARIN's transfer guidance explains that source and recipient submit separate tickets that staff link. An inter-RIR case can have one ticket in each registry and one for each party. A later correction may update the same resource. A merger can move many records without an address-market negotiation.

Counting rows therefore does not necessarily count deals. Counting tickets does not necessarily count requests. Counting addresses gives volume but lets one large block dominate. Counting organisations may merge repeated purchases by a large platform while splitting subsidiaries of one group. Each denominator answers a different question.

ARIN's statistics page provides a revealing example. It reports transfer requests and transfer tickets processed, and explicitly says a processed ticket is a submitted ticket that was closed regardless of final status. It separately reports completed in-region and inter-RIR address volume. That is better than equating every closed ticket with approval, but the public chart does not by itself turn closed tickets into an outcome table with rejection and withdrawal reasons.

A minimum market statement should use a case identifier internal to the participating registries and publish aggregates at four levels. The commercial case is one buyer-seller agreement, if parties report it. The registry case is one requested change of recognised custody for a defined resource set. The party ticket is a source or recipient interaction with one registry. The resource record is each prefix or ASN entry changed.

These levels should reconcile, not be added. One commercial case may generate one registry case, four party tickets and twenty resource rows. The public statement can report all four counts with an explanation. It should never call their sum market volume.

Inter-RIR cases also need a shared case reference. The reference can remain hidden from the transaction log if linking it would expose confidential stages, but both registries need it for aggregate reconciliation. Otherwise each can report its half as a separate request, or one can exclude the case while the other includes it. A bilateral completion must not become two global completions.

Reopened and amended cases require rules. If a buyer changes block size after review, is that one case or a withdrawal and new request? If a seller replaces a corporate document, the case remains one. If parties replace the counterparty after price failure, the commercial case has changed even if the source block has not. Definitions should be set before results are known, with revisions disclosed.

Transparency begins here because institutions can improve apparent performance by changing units. Fewer, larger cases can lower case counts while address volume rises. Splitting a difficult case can inflate closures. A stable observation unit prevents the publication from becoming a story selected by the producer.

Price blindness leaves buyers dependent on private storytellers

Regional Internet Registries commonly say that commercial terms are for the parties. ARIN states that negotiations and financial terms may be resolved directly or through a facilitator. LACNIC says it does not participate in commercial transactions between parties using its listing service. That restraint is appropriate: maintaining an authoritative address record does not give a registry a mandate to set the sale price.

Non-participation, however, is not the same as public ignorance. IPv4 price is already discussed through brokers, auctions, court filings and selected transactions. The problem is coverage and comparability. A broker knows its own deals, not necessarily the whole market. Its published series may be honest and still be shaped by client mix, block size, region, reputation and transactions it wins.

The market significance is visible in APNIC Chief Scientist Geoff Huston's review of addresses through 2025. Using transaction data from Hilco Streambank's IPv4.Global, the analysis reports that prices rose sharply through 2021, later declined, and during 2025 included a low of USD 9 per address for a /14 and a mean of USD 22 per address over the most recent 40 days to 10 January 2026. The article carefully identifies the private data provider. It is a price series, not a census of all transfers.

IPv4.Global's monthly reports add timely observations and block-size commentary. They also come from a market intermediary with an interest in transactions. Another broker may see a different population. Public RIR logs can supply the completed resource denominator, but because they omit price and broker, they cannot show what share of logged volume underlies any private series.

This asymmetry gives large repeat buyers and established brokers an advantage. They observe bids, failed offers, time to close and discounts for restricted routes. A new entrant sees asking prices and selected averages. The information gap can become part of the spread. It also weakens member oversight: if policy delay adds cost, no public series connects price or discount to the registry path.

Requiring every exact contract price in the named log would create serious problems. Consideration can include services, liabilities, tax allocation, clean-up commitments or a portfolio price. Publishing exact terms could reveal strategy, invite collusion or allow a competitor to identify a party through block and date. A registry may lack legal authority to demand the contract and should not pretend a declared number is audited fair value.

The alternative is aggregate price reporting with coverage. Brokers, escrow providers and willing parties submit a standard observation: gross monetary consideration, currency, address count, block-size class, registry pair, contract date band, whether fees are included, whether the deal is bundled, and whether it is an arm's-length permanent transfer. An independent collector validates arithmetic and duplicates, converts currency under a published method, and releases only sufficiently populated cells.

The public learns the distribution and the sample share without learning a named party's price. Registries remain record custodians. Market entities no longer have to accept a private storyteller's sample as the whole market.

Privacy can be protected with separation, delay and thresholds

The usual choice between exact disclosure and secrecy is false. Statistical systems routinely publish useful distributions without releasing individual records. The IPv4 market is thin enough to require stricter design, but not so thin that every aggregate is impossible.

The first protection is separation. The named custody log should not contain an aggregate-cell identifier. The price publication should use quarter, broad block-size class, intra- or inter-RIR path class and perhaps status class. It should not repeat the exact prefix, organisation, transfer timestamp or broker. A reader should not be able to join one table to the other and isolate a contract.

The second is minimum cell size. No price distribution should be released with fewer than five independent commercial cases, and thin directed pairs may require ten. Independence matters: ten prefixes from one portfolio sale are one case. If the threshold is not met, combine adjacent periods or block classes. The publication should say that data were suppressed, not replace the cell with zero.

The third is delay. Quarterly figures can be released after enough time has passed that current negotiation strategy is not exposed. A six-month lag may be justified for very large or unique blocks. Delay should be fixed in advance, not invoked selectively when prices are politically inconvenient.

The fourth is distribution rather than point revelation. Publish the median, 25th and 75th percentiles, address-weighted mean, minimum and maximum only where they cannot identify an outlier, and the number of cases and addresses represented. A median without sample count is weak. An address-weighted mean without a case-weighted median lets one /8 or /12 dominate. Both belong together.

The fifth is coverage. Report the number and address volume of priced cases as a share of recognised arm's-length transfers in the same cell. If only 20% of logged addresses have reported prices, the public can see the limit. If brokers represent overlapping cases, deduplication rules and unresolved uncertainty must be disclosed.

The sixth is controlled access for audit. A trusted reviewer can examine confidential submissions, escrow attestations and duplicate detection without publishing contracts. The reviewer reports error rates, late corrections, provider concentration and whether any contributor could dominate a cell. Contributors can challenge misclassification.

No privacy design is perfect. A unique /8 sale will be visible in the custody log and discussed publicly; its price aggregate may need to remain suppressed. A court filing can reveal a price independently. The standard should protect information it controls, not promise anonymity that public facts defeat.

This approach is more respectful than the current imbalance. Public logs already name source and recipient organisations for many completed transfers. It is hard to argue that even a delayed global median is categorically too private while exact counterparties and prefixes are public. The correct question is which disclosure is necessary for custody and which aggregate is necessary for accountability.

Delay needs a clock for every controller

Price tells entities what scarcity costs. Delay tells them what institutional control costs. Existing transfer dates reveal only the end.

A useful clock starts when the registry has a complete request, not when a user first opens an account or asks an informal question. But the period before completeness also matters because unclear evidence demands can consume weeks. The publication should therefore carry two starting points: first submission and accepted-complete submission. The gap measures intake friction.

From completeness, the case should record stage timestamps: source identity verified; resource eligibility verified; recipient identity verified; need or destination eligibility determined; fees issued and paid; agreement issued and signed; counterpart registry notified; counterpart approval received; cutover scheduled; record changed; technical state checked. Not every registry uses the same sequence, but common milestones can be mapped.

Elapsed time then divides by controller. Registry-controlled time runs while the desk can act. Party-controlled time runs while buyer or seller owes information, money or signature. Counterpart-controlled time runs after a complete bilateral handoff. Joint scheduling time runs while all parties choose a safe cutover. A case can move between states; the totals should reconcile to calendar duration.

Business-day medians are not enough. Publish calendar days because deployment and financing continue over weekends. Show the 25th, 50th, 75th and 90th percentiles and the oldest pending age. Report the share meeting a declared service target. Separate intra-RIR from inter-RIR, IPv4 from AS Numbers, ordinary transfer from merger, and standard from enhanced due diligence.

Pending cases are essential. A registry can make completed cases look faster by finishing easy requests while hard cases remain open. At each quarter end, publish pending age bands: under 15 days, 15 to 30, 31 to 60, 61 to 90, 91 to 180 and over 180, with small-cell protection. Identify the current controller and broad reason.

Time should also be linked to outcome. A rejection after two days and a withdrawal after six months are different burdens. An approval that takes 120 days can destroy a commercial agreement even though the log eventually shows success. Appeals should have their own clocks and should not stop the original duration from being visible.

RIPE NCC openly says inter-RIR cases require more work and take longer because registries coordinate across time zones and synchronise updates. That is a plausible hypothesis. Stage data would show whether the extra time lies in necessary cutover, duplicated review or unowned handoff. ARIN says duration can vary because several actors are involved. Controller clocks would identify which variation members can ask ARIN to improve.

Good time reporting does not punish careful review. It lets an institution show that a long case spent 80 days waiting for a disputed corporate succession rather than in an analyst queue. It also prevents "complexity" from becoming an explanation that cannot be tested.

Rejections and withdrawals are the missing denominator

A market cannot be transparent if it records only recognised transfers. Institutional accountability cannot be credible if it records only approvals.

The minimum quarterly reconciliation is simple in form:

opening pending cases + new complete cases + reopened cases = approvals + rejections + withdrawals + expiries + merged or replaced cases + closing pending cases

Each term needs a definition. A rejection is a registry decision that the request does not satisfy a stated requirement. A withdrawal is a party's decision to stop before final determination. An expiry is an administratively closed case after notice and a stated response period. A merged or replaced case is not a successful outcome; it is a bookkeeping movement that must point to the successor case internally so it is not counted twice.

Reasons should be mutually exclusive at the primary level and permit secondary tags. Primary categories can include holder identity not established, signatory authority not established, active dispute, resource ineligible, holding period, recipient identity, demonstrated need, regional nexus, counterpart incompatibility, sanctions or legal order, fees or agreement incomplete, duplicate request, technical cutover failure and party commercial withdrawal. "Other" should be rare, reviewed and explained in aggregate.

The category describes the decision, not the applicant's character. A need shortfall is not fraud. Missing documents are not proof that the holder lacked rights. A withdrawal after a price change is not a registry rejection. Keeping these distinctions protects parties and improves analysis.

Withdrawals need a limited voluntary reason menu because they often reveal institutional effect before a decision. The parties can choose commercial terms failed, financing failed, counterparty changed, timeline exceeded, policy eligibility concern, technical concern or reason withheld. Refusal to answer should remain available. The registry should not infer a reason from correspondence.

Pre-filing attrition belongs in a separate sample. Listing services, qualified facilitators and brokers can report how many inquiries were screened out for path incompatibility, needs evidence, block reputation, size, price or seller authority. Those numbers cannot be added to complete registry requests because populations overlap. Publish them as surveyed funnel data with contributor count and coverage.

Appeals complete the picture. Report how many rejections were challenged, affirmed, reversed, remitted for more review or still pending, plus time to disposition. Suppress thin categories where identity could be inferred. A high reversal rate may reveal unclear standards; a zero appeal rate may mean predictable decisions or inaccessible review. The statistic requires context, not celebration.

ARIN's processed-ticket chart already demonstrates that institutions can count closures irrespective of final status. The next step is to disclose the distribution of those statuses in a case-consistent form. LACNIC's policy already mandates a completed transfer log. Adding an aggregate decision table does not expose the legal documents the parties filed.

The denominator will not answer every fairness question. Brokers may deter weak cases, and sophisticated applicants may file more complete evidence. It will end the current condition in which a registry can point to a clean list of successes while the public cannot see whom its rules excluded.

The minimum public statement needs five linked panels

A practical standard should be small enough to publish every quarter and rich enough to resist institutional storytelling. Five panels meet that test.

Panel one: completed custody changes. Keep the existing cumulative transaction log: resource set, source and recipient organisations where lawful, source and destination registries, transfer date, broad type and corrections. Add a stable revision note and data-freshness timestamp. This remains the authoritative public history, not a price table.

Panel two: case outcomes. For each registry, directed inter-RIR pair, resource class and case type, publish opening pending, new complete, approved, rejected, withdrawn, expired, replaced and closing pending cases. Include address volume separately from case count. Reconcile the panel arithmetically.

Panel three: elapsed time. Publish first-submission-to-complete intake time; complete-to-decision time; decision-to-cutover time; total calendar time; registry-, party-, counterpart- and joint-controlled days; percentiles; target compliance; and pending-age bands. Complex enhanced-review cases can be a separate class, but the rule for entering that class must be public.

Panel four: reasons and review. Publish primary rejection, withdrawal and pending reasons, missing-evidence categories, appeal filings, outcomes and time. Link each decision category to the policy clause or operating rule it represents. Show corrections where a classification changed after review.

Panel five: market aggregates. For permanent arm's-length IPv4 transfers with reportable consideration, publish case and address coverage; block-size bands; intra- and inter-RIR classes; price currency method; median; quartiles; case-weighted and address-weighted measures; dispersion; broker or direct share where sufficiently populated; and the lag between contract and record dates. Suppress cells below the privacy threshold.

Every panel needs metadata written for readers: reporting period, producer, extraction date, definitions, inclusions, exclusions, known gaps, revisions and contact for correction. A public cryptographic digest of each release would make later alteration visible. Dated snapshots should remain available.

The panels should not be collapsed into a composite score. A cheap market can be slow. A fast desk can reject many applicants. A high approval rate can reflect severe broker screening. A rising price can reflect scarcity, block mix or changing demand rather than registry performance. Keeping dimensions separate forces the conclusion to match the evidence.

Nor should a global total erase pair differences. Inter-RIR friction is directional. Publish ARIN-to-APNIC separately from APNIC-to-ARIN where cell size permits. If privacy requires combining, show which routes were combined and why. A global median dominated by RIPE NCC internal transfers says little about a small operator trying to move space into LACNIC.

The minimum is not technically extravagant. Registries already hold case states and dates needed to administer requests. Brokers and escrow providers already hold transaction amounts for their own cases. The challenge is institutional: agree definitions, preserve confidentiality, accept reconciliation and publish the failures that make performance less flattering.

Data quality needs revision history, not silent correction

A cumulative file can improve as errors are found, but improvement without revision evidence can rewrite the past. Transfer data are especially vulnerable because organisation names change, blocks are split, mergers are reclassified and historic records are imported from older systems.

Each feed should identify production time, covered interval and format version, as the NRO design already anticipates. It should also carry a release-level count by resource and type, the prior release digest, and a correction file stating which records were added, changed or removed and why. Personal or security-sensitive reasons can be broad; the fact of change should not disappear.

Freshness is part of quality. A file named "latest" encourages readers to assume current coverage. The five live RIR feeds inspected for this article did not share the same newest transfer date. That may reflect different market activity, production schedules or stale publication. A freshness panel should distinguish "no new transfers" from "file not regenerated" and "generation failed." A signed zero-activity statement is data.

Cross-RIR transfers should reconcile on both sides. The source and recipient feeds should agree on resource set, transfer date, counterpart and type. A public reconciliation report can list unmatched records after a grace period without exposing internal tickets. Differences caused by time zones, delayed publication or format mapping should be classified and resolved.

Names need careful treatment. The organisation named in a public registry may be a legal entity, trading name, parent, subsidiary or historical holder. A name change should not look like a new sale. Stable public organisation identifiers would improve research, but they must not expose private account IDs. The identifier should survive spelling corrections and link to public RDAP identity where lawful.

Price aggregates have their own revision risks. Late broker submissions can move a median. Currency conversions can be corrected. A portfolio may be split after audit. Every release should preserve the originally published value, revised value, reason, affected cell and date. Silent replacement would let an institution or contributor smooth an inconvenient market movement.

Independent validation should sample from each panel. For custody, compare the feed with public registry state and bilateral counterparts. For cases, reconcile opening and closing balances. For time, recompute totals from event timestamps. For price, verify a confidential sample against escrow or settlement evidence and test duplicate detection. The auditor reports method and error without publishing contracts.

Data quality also limits conclusions. An exact-looking mean from 30% coverage is not a market price. A rejection rate excluding incomplete intake is not an application rate. A transfer count mixing mergers and sales is not turnover. The publication should place those warnings beside the figure, not in an obscure note.

Trust does not require error-free data. It requires errors to be discoverable, corrected and preserved as part of the record.

Brokers can contribute evidence without owning the market narrative

Brokers occupy an awkward but necessary position. They see offers that never reach registries, negotiate prices, understand block reputation and learn which institutional paths close reliably. They also compete for business and can benefit from a particular account of supply, demand and urgency.

Excluding them would discard the best available price and attrition observations. Allowing one broker's series to stand for the market would exchange public opacity for private concentration. The standard should use plural contributors and disclose concentration.

Each participating broker can submit signed quarterly records to an independent collector. The record identifies a pseudonymous commercial case, block-size band, address count, source and destination RIR, contract-date month, completion-date month, currency, gross consideration, included services, broker fee treatment, escrow status, outcome and reason if the deal failed. Exact party names and prefixes can be retained under confidentiality for deduplication and audit, then excluded from public aggregates.

Direct deals need a channel too. Buyers, sellers, bankruptcy administrators and escrow providers can submit the same form. Where two parties report the same case, the collector matches it and resolves discrepancies rather than counting both. Contributors receive a receipt and can later correct a fact with an audit trail.

Coverage reporting should name provider concentration only at a safe aggregate. If one broker supplies 80% of priced cases, readers need to know that the series largely represents one venue. The broker need not reveal its client list. An independent reviewer can confirm the share.

Listing data should remain separate from completed price. An asking price is not a sale. A bid is not a sale. A broker's valuation is not a sale. A court-approved amount may bundle assets. The public panel can show median ask and bid spreads as different series if definitions and coverage are clear, but neither should be merged into realised consideration.

Registries can assist without demanding contracts. They can let parties opt to send a completion token to the collector, confirming that a recognised transfer occurred for a pseudonymous case and address count. The token need not carry price. It allows the collector to distinguish completed recognised transfers from unrecorded operational leases or abandoned deals.

This division respects institutional roles. Brokers provide commercial observations. Escrow providers attest settlement where authorised. Registries attest custody change. An auditor tests the aggregation. Members and operators see coverage and method. No single actor can claim that its partial view is the entire market.

The design also leaves room to study unrecognised movement. Researchers can compare routing, RDAP and named transfer records, as prior academic work has done. Such inference should be labelled probabilistic and should never be inserted into the authoritative custody log as a proven sale. Market transparency improves when evidence classes remain separate.

NRS should publish the comparison, not sell permission

Number Resource Society can make this architecture practical if it treats transparency as a ledger service. It need not become a broker, receive title documents for every transfer or persuade registries to hand it approval power.

Its first role is definitional. NRS can publish the case taxonomy, stage clocks, reason codes, privacy thresholds, currency method, revision rules and coverage formula under open change control. Each definition should be usable by any RIR, broker or researcher. Competing implementations should produce the same quarterly reconciliation from the same events.

Its second role is aggregation. NRS can receive signed statistical returns from registries, brokers, escrow providers and parties; deduplicate commercial cases under confidentiality; and publish the five panels. Raw identity evidence can remain with the submitting institution where possible. NRS needs only enough encrypted linkage to test duplicates and audit coverage.

Its third role is portability. A buyer or seller should be able to export its own timestamps, submissions, decisions and correction history in a standard packet. If a registry misclassifies party delay or loses a document, the party has a verifiable record. Portability reduces dependence on a private ticket history and makes independent review possible.

Its fourth role is comparative restraint. NRS can state that one registry reports 90th-percentile completion while another publishes only an average. It can identify missing rejection denominators and stale releases. It should not rank institutions with a single score that rewards selective reporting. The comparison is strongest when it preserves gaps.

The limit is as important as the service. NRS must not certify a "fair price," require its report before a registry transfer, direct parties to a preferred broker or sell faster recognition. It should not infer consent to public price disclosure from membership. Contributors choose defined confidential terms, and aggregates meet a public-interest standard independent of sponsorship.

Funding must not determine the narrative. A large broker can pay for data engineering without controlling definitions or release timing. A registry can correct an error without vetoing an unfavourable result. An operator can challenge its own classified delay without obtaining another party's documents. Governance records should disclose funding and conflicts.

This is the positive NRS proposition in concrete form: pay for a minimal, comparable ledger; make evidence portable; show where the public record ends; and refuse to convert information service into gatekeeping rent. An institution demonstrates legitimacy not by knowing every contract but by making the limits of its knowledge measurable.

If NRS can be replaced by another collector using the same open definitions and signed inputs, the design has succeeded. Transparency should not depend on a new monopoly.

Registries could publish the first accountable version now

The complete standard will take coordination, but the absence of perfection is not a reason to preserve the current blind spot. Each RIR can publish a first quarterly statement from data it already controls.

Start with case reconciliation, outcome categories, elapsed-time percentiles and pending age. Do not wait for price. These fields test the transfer desk itself and do not require a registry to enter commercial negotiations. Publish definitions, exclusions and a one-year back series if reliable. Where historical event timestamps are incomplete, say so and begin prospectively.

Next, align directed inter-RIR cases. Each pair can reconcile complete and pending cases, handoff time and unmatched records. The institutions already communicate to change custody. A public aggregate agreement should be less sensitive than the documents exchanged in the case.

Third, add a voluntary confidential price pilot with at least three independent brokers, direct-party access and an auditor. Release only cells meeting threshold and publish coverage. A pilot result with 25% coverage is more useful than an unsupported claim of the global price, provided the limitation is prominent.

Fourth, establish correction and appeal reporting. Every quarterly release should list changes to prior periods. Every decision reason should map to a current policy clause. Members can then see whether a policy produces recurring confusion or reversals.

Fifth, commission an external data-quality review after four quarters. The reviewer should test case units, bilateral duplicates, pause attribution, privacy leakage and price concentration. Recommendations and management responses should be public.

The first release will look less polished than today's success log because it will contain gaps, pending cases and suppressed cells. That is a sign of improved honesty. A mature institution does not need every chart to flatter it.

Members should insist on publication because they fund the transfer desks and live under the rules. Buyers and sellers should support it because predictability lowers closing risk. Brokers should support it because a trusted aggregate can expand participation while distinguishing credible contributors from unsupported advertising. Operators should support it because accurate custody and timely technical transition affect routing safety.

Price is not the only missing field, and it should not delay the rest. A market can become more accountable before it becomes fully priced. But the final ambition must include price coverage. Otherwise institutions will know which resources moved while only private repeat players know what access cost.

The immediate test is modest: can each registry explain, for the last quarter, how many complete cases it opened, approved, rejected, lost, carried forward and how long each stage took? If the answer is no, a list of completed blocks cannot reasonably be called transparent market oversight.

A transparent market records the unseen queue

The transfer log should be defended for what it does. It preserves a public history of recognised changes, supports routing and provenance research, and gives counterparties a way to verify that the ledger no longer names the old holder. Removing names or making the file less accessible would not create privacy-respecting accountability. It would destroy useful evidence.

The reform is additive. Keep the custody record. Add a statistically protected account of the market and the institution.

That second account must include the unseen queue: complete requests awaiting action, cases paused for parties, bilateral handoffs, rejections, withdrawals, appeals and inquiries screened out before filing. It must show the price sample and its limits. It must distinguish a commercial case from a ticket, a prefix row from a deal, a merger from a sale and a transfer date from a contract date.

These distinctions prevent easy claims. A rising transfer count is not proof of greater access. A high approval rate is not proof of fairness if brokers pre-screen most applicants. A low median is not proof of speed if old cases remain pending. A private broker average is not the global price if coverage is unknown. An exact price is not comparable if it bundles services or comes from a different block class.

The minimum public fields are therefore not administrative decoration. They define what members are able to know about a power exercised in their name. A registry that can reject a transfer, demand more evidence or coordinate a delayed cutover should be able to report how often, why and for how long, even when individual documents remain confidential.

Price aggregation completes the accountability boundary. It does not ask the registry to endorse commodification or set a rate. It shows the economic environment in which registry decisions operate. If a policy closes a route, prolongs escrow or favours buyers able to document conventional demand, price and delay data may reveal the incidence. Without them, distributional effects remain anecdotal.

NRS points toward a better institutional form because it can join these observations without owning the underlying rights. A minimal ledger service authenticates records, preserves portability and publishes comparable evidence. It does not decide whether an address is worth USD 20 or whether a buyer's business plan deserves approval. It makes those separate claims visible.

The public transfer row is the end of one story. Market transparency begins by recording all the cases that did not reach it.

Sources