Summary

  • Mueller and his co-authors reconstructed an emerging IPv4 transfer market from records that no single registry or public database captured completely.
  • Their method made transactions and route visibility easier to examine, but a recorded handoff or absent BGP announcement did not settle price, business need or service continuity.

When a scarce resource changes hands, the first analytical problem is often not valuation. It is deciding what counts as a change of hands, and which records can establish it. That was the awkward task confronting researchers studying IPv4 address transfers in the early 2010s: the market was controversial, the regional registries disclosed different slices of it, and the public debate often treated IPv6 as if it made the question disappear.

Milton L. Mueller had approached the issue as institutional economics before it became a data exercise. His 2008 paper examined proposed transfer policies and the incentives facing organizations that held address blocks and organizations seeking them. Four years later, with Brenden Kuerbis and Hadi Asghari, he helped turn the policy argument into an empirical study. “Dimensioning the Elephant” did not start with one authoritative transaction database. Its authors assembled bankruptcy-court records, regional Internet registry statistics, Whois and historical Whois records, and lists of transferred blocks. The paper’s own metaphor acknowledged the problem: each source showed only part of the market. The 2008 analysis and the 2012 study mark a shift from arguing about whether a market could exist to testing what its records could show.

The records did not line up neatly. In the study’s historical snapshot, RIPE NCC did not publish a public log of transfer events. ARIN’s list identified address blocks processed under a policy, but did not itself name buyer and seller, date, or price; the researchers had to combine it with Whois and “Whowas” records. APNIC’s log gave more detail about releasing and acquiring organizations and effective dates, yet combined commercial transfers with moves caused by mergers, acquisitions or internal restructuring. The authors tried to separate those cases and admitted that some classifications remained uncertain.

These are descriptions of the sources as they stood in that study, not claims about present-day registry practice.

That unevenness shaped what “the market” meant in the analysis. A transaction, a transferred block and an individual address are three different units. One agreement can cover several prefixes; one organization can move blocks in more than one deal; and prefixes vary greatly in size. The authors grouped some transfers involving the same parties on the same date as one transaction while noting that more complex multi-party deals could defeat the rule. Their spreadsheet therefore made assumptions visible instead of pretending that an administrative list was already a complete market dataset.

A second evidence layer asked whether blocks appeared in global routing data before and after a transfer. Asghari’s PyASN tool queried /24 address ranges against archived RouteViews BGP dumps at six-month intervals from January 2010 through July 2012. That gave the authors a repeatable way to observe whether a prefix had a visible origin in sampled routing tables. It was not a traffic meter, a census of customers, or proof of what a holder planned to do. A prefix absent from a particular snapshot could be unannounced, filtered, reserved, used privately, or simply outside the observation window.

The Nortel-to-Microsoft bankruptcy sale showed both the value and the limit of this approach. The paper examined 38 transferred blocks and reported that by July 2012 it could observe seven, totaling 10,496 addresses, routed by Microsoft. It treated the modest routed share as relevant to a short-horizon “needs assessment.” But a route snapshot could not tell the whole economic story. A buyer might value an address block as an option against future scarcity, or need it on a planning horizon different from the registry’s test.

Conversely, a court-approved sale and a registry record alone could not prove that a network used the resource or that an operational transition was smooth. The study itself described disputes over the ARIN review and nondisclosure of the special contract’s terms.

Later work on transfer detection sharpened the warning. Ioana Livadariu and co-authors found that inferring transfers from changes in BGP origin can generate many false positives because routing changes have causes other than a sale. That is a useful check on any temptation to turn network visibility into an ownership ledger. The reverse inference is also unsafe: a recorded transfer does not establish continuous service, and a missing route does not establish that an address was unnecessary.

Mueller’s contribution here is not a final census. It is a research design that links institutional records to network observations while keeping their meanings separate. The follow-on iConference paper, “Buying Numbers,” framed those assembled data around policy questions such as needs assessment and property rights; the University of Illinois repository preserves its abstract and publication record. Georgia Tech’s profile places the work within Mueller’s longer research on the political economy of Internet governance and identifies him as a professor and director of the Internet Governance Project.

The biographical record explains why the market was a natural subject for him; the study’s method shows how he made it testable.

The distinction remains useful for operators and policymakers: a registry can record a transfer, a court can document a sale, and a routing table can show a prefix announcement, yet those are separate observations. Joining them improves the picture. Treating them as interchangeable would make the picture less reliable.

Sources