There's talk that in the battle between the USA and Europe over control of ICANN, which may come to a head at the upcoming World Summit on the Information Society in Tunis, people will seriously consider "splitting the root" of DNS.

I've written a fair bit about how DNS works and how the true power over how names get looked up actually resides with hundreds of thousands of individual site administrators. However, there is a natural monopoly in the root. All those site admins really have to all do the same thing, or you get a lot of problems, which takes away most of that power.

Still, this is an interesting power struggle. If a large group of admins decided to switch to a new DNS root, different from ICANN, they could. The cooperation of Microsoft, which includes the default root list for IIS, and Paul Vixie, who puts that list in BIND, would play a large part in that as well.

In fact, many times in the past people have split the root by creating alternate, "superset" roots which mirror the existing .com/.org/.net/etc. and add new top level domains. Some of these have been "innocent" efforts frustrated at how slowly ICANN had created new TLDs, but in truth all of them have also been landgrabs, hoping to get ownership of more generic terms, furthering the mistake that was made with .com. ICANN is also furthering the mistake, just more slowly. (The mistake is ignoring what trademark law has known for centuries -- you don't grant ownership rights in ordinary generic terms.)

All of these superset attempts have also failed. I don't think I have ever seen anybody promote a URL using one of the alternate root TLDs, or give me an email address from an alternate root TLD. I consider that failure.

This is, of course, what creates the natural monopoly. Few people are interested in setting it up so that two different people looking for a domain get different results. That applies to the fact that most people get an error for www.drug.shop (in the new.net alternate TLDs) and a few get the registrant's site, but it applies even moreso to the idea that Americans would get one answer for foo.com and Europeans a different one.

Because of this, Larry Lessig recently suggested he wasn't worried about a root split because there would be such strong pressure to keep them consistent.

The difficulty is, what's the point of creating your own root if you can't actually make it any different from the original? The whole point of wanting control is to have your way when there is a dispute, and to have your way does not mean just doing it the same as everybody else lest we get inconsistent results.

It's possible that a group of nations might try to wrest control in order to do nothing at first, but eventually create a superset of TLDs which would, for the first time, be a success. That might work, since if all the nations of the world except the USA were to go to a new root set, it would be hard for the private individuals in the USA who control name servers not to follow. But then the new group would no doubt attempt at some point to issue policies for the existing top-level-domains and country code domains.

People do want their own way, and that's bad. The truth is that nobody should be in control of the root. Not the USA, not Europe, not the UN. It is actually possible to build a root under the control of no nation. Instead, it could be under the control of a committee dedicated to a set of very simple principles. Principles so simple that disputes over them are few, and deviation from the principles is obvious. That means all real disputes must be pushed down a level, to where specific, actionable entities are in control, and that's fine if there are lots of them, and they are truly competing -- on policy and jurisdiction, and not just on price as we have today.

Since this plan calls for the eventual slow wasting away of the generic TLDs (over which we are doomed to have global and local battles) I've refined my proposal there. There need be no fee for new, non-generic TLDs, and the fee for the generics should go up, exponentially. This year 25 cents, then 50 cents, a buck, 2 bucks and so on. Gives everybody plenty of time to migrate away from .com, but also makes very clear to even the very richest that they can't sustain ownership of terms in the generic space forever. (In 20 years the fee would be over $250,000 per year) After everybody migrates away, the non-generics would have to pay minimal fees instead of getting subsidies.

ICANN could even allow this plan to start by allocating one of the rare single letter top-level-domains (it doesn't matter much which) to a trans-national group. Domains of the form yourname.{coinedterm}.i would be almost as saleable as yourname.{coinedterm}.