ICANN has released to the public a report they commissioned called “Revisiting Vertical Separation of Registries and Registrars,” written by CRA International. It is being referred to as the “CRAI Report.” Readers in the U.S. and the U.K. may not know it, but most top-level (TLD) domains in the world don’t have registrars—you go straight to the registry and buy your domains from the source. Only in among generic TLDs and selected country code TLDs are there registrars.

ICANN was founded to do two things: promote competition among existing gTLDs (i.e., lower the price), and create competition among TLDs (i.e., reduce the monopoly position of .com). The latter goal is finally being attempted with the latest round of new TLDs, while reduction in price was achieved almost immediately by means of interposing a registrar between the end-user and the registry. Under this scheme, a registrar, Tucows for example, would buy a .com from VeriSign for $6, and resell it to the customer at $10. Another registrar might add some hand-holding and sell it for $20. The point was, customers could choose based on price and service. Prices went from $35/yr to about $10, where they remain.

So the registry-registrar split is a good thing, right? Maybe. The CRAI Report looks at whether ICANN should keep that structure in place for all new TLDs, or loosen the strictures a little. It starts with a capsule history of the reasons for separating the registry and registrar functions, and then looks at the advantages and disadvantages. On the whole, they come down in favor of retaining the division.

I’m not so sure. When all the world was cornflakes and cheerios (.COM and .NET), it made sense that there shouldn’t be a monopoly provider. Now that we’re about to have everything from frosted flakes to granola to muesli to coco-krispies (all the new flavors of TLDs), I’m not sure that it makes that much sense any more.

Consider these possible applications (imaginary but entirely possible):

• .HOPI, restricted to members of the American Indian Hopi tribe. Given the small numbers of the tribe and low availability of Internet services on the Hopi reservation, very few registrations are expected. Nonetheless the tribe has determined that its future requires an identity on the Internet, and is prepared to underwrite the expense of a new TLD. The tribe has also decided that it needs a bilingual presentation of the registration activity to help educate its members about domains and the Internet in general. BUT no registrar is at all interested in offering .HOPI, and certainly not bilingually, and especially doesn’t want to clutter up its valuable web real estate with any educational content.
• .ROOTS, restricted to residents of the new world who trace their ancestry back to Africa. In this case, the potential number of registrants is quite large, but there is a qualification procedure, which might involve several steps, including examining documentation. Again, no registrar is going to want to set up the necessary workflow.
• .CAIRO, wanting to reach the highest score on ICANN’s comparative evaluation scale by being as restrictive as possible (a criterion ICANN needs to fix), has determined that domain names should go only to Cairo residents who voted in the last (sham) election for the ruling party. Again, registrars are unlikely to be in the least interested. Not only do documents have to be examined, and not only is this a non-English language, it’s also an IDN implementation. Furthermore (although I admit this is unlikely), some registrars might actually have scruples about propping up a dictatorship.

I could invent examples for days. There are many business models for registries that almost require a dedicated registrar, or (better) a registry that can sell directly to the public. These are all cases where a registry is going to have a lot of trouble attracting a registrar to sell domain names. At present, there is no obvious solution for such registries, except by skirting ICANN’s cross-ownership rules (a registry can own no more than 15% of a registrar) by purchasing or starting a registrar via an “arms-length” transaction—basically, by breaking the rules.

Cries of “monopoly” are misplaced. In your supermarket aisle, Shredded Wheat has a monopoly on shredded wheat, Lucky Charms has a monopoly on lucky charms, Corn Pops has a monopoly on corn pops, Fruit Loops has a monopoly on fruit loops, Grape Nuts has a monopoly on grape nuts, etc. etc. We don’t live in a world where there is only one cold cereal, sold by many vendors; instead, we have a wide variety of cold cereals, each sold by a “monopoly”—also known as a “brand.” This same regime will inevitably apply to new TLDs as a natural consequence of the diversity in TLDs that ICANN’s process is designed to achieve. ICANN needs to realize that the rules of yesterday might not apply to tomorrow.

Probably the best solution, and where ICANN will end up eventually, is that .com, .net, and the other gTLDs that have found acceptance among registrars, should retain the registry-registrar split. Others, particularly those labeled as “community” TLDs by ICANN, should be able to choose to use the ICANN registrar channel or not. Failing that, either registrars need to agree to offer all new TLDs according to the needs of that registry (highly unlikely), or the rule that a registry cannot own more than 15% of a registrar should be abolished.