At the Seoul meeting the Registrar Stakeholder Group has adopted a position paper supporting the removal of rules that prohibit them from selling names in TLD registries they own or control. Prior to this, many observers of the controversy assumed that it was only Demand Media (ENom) and a few other large registrars who supported the removal of this restriction from new TLDs. Now we see more of an industry-wide cleavage between registrars and registries.
What motivates Registrar advocacy of this position is the desire to more effectively market and promote new TLDs they would like to start. It may also enhance their ability to monetize the value of second-level domains in new TLDs. Without these capabilities, the business case for new TLDs may not look that good.
What motivates Registry opposition to this change is that it makes their registry back-end services less competitive when vying for the business of new TLDs. The influx of hundreds of new TLDs could be a boon to existing registries such as Afilias or Neustar, because few of them will operate their own infrastructure. But registrars can enter this back-end registry market also. Larger, established registrars can take most of this business away from incumbent registries if they are allowed to sell names under the TLDs they serve. That is because incumbent registries do not have the marketing channels to the public that the registrars have. Thus, a start-up TLD would prefer to work with a registry operator that can also distribute the names. Also, incumbent registries oppose the change because they fear that it will incent registrars to promote their own new TLDs at the expense of their existing ones.
In this controversy there is some terminological confusion about “vertical integration.” This debate is not really about vertical integration, at least not yet. Full vertical integration of the Registry-Registrar functions would mean that the same corporate entity would bring together technically the registry and registrar functions into the same operational home. At its most extreme, vertical integration would mean no functional separation and no equal, nondiscriminatory access to the registry by competing registrars. The efficiency benefits of such integration cited in studies by ICANN's consulting economists would not materialize unless there was full vertical integration of this sort. What does get integrated, however, is the information about domain names that flows through the registry along with the marketing and promotional strategies of the registrar.
The registrar position statement supports equal access to a registry by all registrars, and requires registrars to remain functionally separated from the registry. Thus, it does not advocate vertical integration. It only advocates the removal of the ban on retail distribution of a registry's own names. This is more than just “cross-ownership,” however; it involves eliminating one of the enforcement mechanisms meant to support a level competition among registrars. And longer term, it could lead to more extensive integration.
The public policy tradeoff is a tough one. There is no doubt in my mind that the change advocated by the registrars will promote additional competitive entry in the registry market. If it were just about that, it would be simple. There are two complications, however. One is whether the particular type of competition fostered by removing this restriction would actually benefit consumers in this unusual type of market, where only one customer can occupy the product (a unique name). None of the economic studies commissioned by ICANN deal with that question adequately. The other is the asymmetrical nature of the change; i.e., it allows registrars some new options but does not provide the same to incumbent registries. It would be possible to make the policy symmetrical (i.e., allow registries without market power to sell direct to customers, too), but once you start going that you need to review the assumptions and policies underlying the vertical separation policies more comprehensively. And that means that the whole thing needs to go through a policy process, which could delay the advent of new gTLDs for two years or more.
So it's a tough choice: it's distasteful to thwart competition in the near term in order to develop a more comprehensive policy later; on the other hand if we don't work out the more comprehensive policy we may end up with new problems and distortions. I'll write more later about how to respond to this policy conundrum.