Summary

  • ICANN's original AGP history describes a typically five-day window for correcting domain registrations. Bulk domain tasting used that same refund to test names for traffic and advertising revenue, keeping winners and returning losers.
  • ICANN's August 2009 status report separates a temporary US$0.20 charge on excess deletions from a later policy that withheld the registry refund above a monthly allowance. It records a 99.7% decline in AGP deletions between June 2008 and April 2009 across TLDs offering that grace period.

A refund became a business model

A five-day cancellation window makes sense when a person mistypes a domain name. The registry can reverse the registration, credit the registrar, and let a mistake be corrected cheaply. ICANN's history of the Add Grace Period says the same mechanism served error correction and some fraudulent-registration remedies. It also supplied a different kind of customer with an attractive option. Register many names, put them to work briefly, keep those likely to earn, and return the rest before the charge sticks. The buyer gets a test; the system processes the churn.

The policy question was therefore more exact than “should deletions be banned?” A blanket end to the grace period would also make genuine mistakes costlier. The institutional target was the cost-free excess. The GNSO's proposal adopted by the ICANN Board on 26 June 2008 drew a monthly line for each registrar: the greater of 10% of its net new registrations or 50 names. Above that line, a registry operating the grace period could no longer refund the registrar's registration fee under the eventual consensus policy. The decision attached a price to scale while leaving room for routine correction.

Twenty cents first; the registration fee next

ICANN did not wait for the final policy to take effect. Its status report describes a separate FY2009 budget provision running from 1 July 2008 to 30 June 2009. It treated deletions above the same monthly threshold as transactions subject to ICANN's US$0.20 fee. The fee was small beside a registration price, but it removed the completely free return for large volumes. The Board approved the budget and the policy recommendation at its June 2008 meeting as distinct acts. A vote to adopt a policy was not its effective date.

ICANN announced implementation documents on 17 December 2008; its consensus-policy list dates the AGP Limits Policy's effect to 1 April 2009. The new rule used the same allowance but changed the economic consequence. Excess deletions lost the registry's refund, rather than merely attracting a twenty-cent ICANN fee. The registrar could request an exemption for extraordinary circumstances with documentation. That exception kept a path for unusual legitimate events; it was not a general licence for a rolling free trial.

ICANN supplied an unusually helpful worked example. A registrar makes 1,000 new registrations and deletes 300 within the grace period. Net new registrations are 700, so 10% permits 70 refundable deletes. The remaining 230 would have cost nothing before the budget provision, US$46 under the twenty-cent measure, and US$1,552.50 under the policy if each lost refund were US$6.75. Those figures are ICANN's illustration; the actual withheld refund depends on the registry's price. The arithmetic exposes the decision. A free option becomes less interesting when every losing ticket has a price.

The collapse was real; the unit matters

The August 2009 report says AGP deletions across TLDs offering the grace period fell 99.7% from June 2008 to April 2009. Those endpoints span both the interim fee and the first month of the permanent policy. The measured unit is a deletion during the grace period, not a psychological reading of why each name was registered. The change strongly supports the case that the price of speculative churn mattered. It does not allocate the entire 99.7% to the April policy alone.

ICANN continued monitoring. Its 2010 announcement of a proposed final report said results had persisted and that the GNSO Council saw no need for further policy work at that point. That is ICANN's institutional assessment, backed by its own registry reports. The more revealing design fact remains the sequence: the registrar's free return acquired a modest cost, then a much larger loss of refund above a carefully defined allowance. The grace period survived; the mass free option did not.