On 3-Dec-07, at 7:39 AM, Danny Younger wrote:
If the Add Grace Period is not eliminated, then registrars will simply adjust their business models to compensate for any registrar level transaction fee that might be imposed, and domain tasting will continue.
I'm not sure this is entirely correct. Levying a transaction fee for all registrations, regardless of their duration, will change the economics of the transaction. The effect of a transaction fee at $0.01 will be different than a transaction fee set at $6.42. A transaction fee of $0.01 requires that 1 in 642 tasted names yield more than $6.42 per year, a transaction fee of $6.42 (which is essentially equal to eliminating the AGP) requires that every name tasted yields more than $6.42. In other words, it is possible to properly modify the fees associated with a registration such that it becomes unprofitable for tasters to guess which names might run a profit over the course of a year. Levying a $.20 fee on AGP transactions would mean that 1 in 32.1 names needs to yield more than $6.42. In April of 2006, the tasting keep rate was estimated to be approximately 1 in 17. In February of 2007, this was estimated to be 1 in 13.3. Despite the apparent increase in the take rate over this period, I'm not sure that it reflects anything more than increased sophistication in the tasting and monetization models. [as you read the rest, please remind yourself that I am not a mathematician and the actual numbers might be incorrect...] If these numbers are accurate, then the economic solution to the problem is simply to levy a fee that makes the odds exceptionally bad for tasters - i.e. if they know that 1 in 13.3 names will be profitable, but that finding that one name may cost them $9.97 (reflecting a $.75 levy), then they are highly unlikely to engage in the behavior -i.e. it would likely be unprofitable. Without even pulling my stats on cancellations, I can definitely say that I'd be happy to pay a $.75 fee for mistakes. Systems testing on the other hand might require some special treatment - for example, the registry could provide the registrar with a string that they would use for testing - something like systemtestingdomain0001-IANARID-69.com, where the numeric increases by 1 for each subsequent name ran through the tests. All of the names tested would also need to automatically expire after some reasonable period of time (7 days, 10 days?) after the testing occured and could not be re-registered by anyone. If a registrar ran subsequent tests, they would simply register the next name in the sequence from where they left off in their last testing.... Just some ideas... -r