At the December 2004 ICANN meeting in Cape Town, Vint Cerf said this to the Public Forum:

“I want to go on record as saying… that I am no longer sure that I have a strong understanding of why I would be motivated to create a new TLD… for many years we didn’t create any new ones. And it wasn’t because we didn’t have an apparatus for doing it. It was because it wasn’t clear what the rationale was for creating new TLDs. If we pick the wrong philosophical basis and we try to codify that and we end up creating so many TLDs that we actually create a problem with the domain name system, we have a problem.” (Link)

Dr Cerf posed a question that has yet to be answered or even discussed by the DNS stakeholder community—while the technical and business cases for the introduction of new TLDs have been successfully made, what is the philosophical case for adding new TLDs? What semantics are encoded in TLDs, and how could those semantics be expanded in a consistent way?

The Purpose of TLDs

The current stock of TLDs can be categorised into two groups:

1. TLDs that identify an entity based on geographical location: ccTLDs, .EU, and potentially .ASIA in the future
2. TLDs that identify an entity based on “type”: eg, commercial (.COM), educational (.EDU), military (.MIL)

Furthermore, some ccTLD operators use subdomains to add additional semantics: .UK is divided into a range of second-level domains, including .CO.UK for commercial organisations, .AC.UK for academic organisations, .SCH.UK for schools, and .ME.UK for private individuals.

A simple pattern for generic and sponsored TLDs can be construed from looking at the TLDs currently in use: the purpose of a domain name suffix is to identify the type of organisation, and/or the geographic location of that organisation.

It is my contention that the when the DNS stakeholder community considers the introduction of new TLDs, that we should try to follow this pattern, and not break it without good reason.

Adding new TLDs

If we’re to consider new TLDs for entry into the root zone, what criteria should we look for?

1. Geographic TLDs

While the ISO3166 list provides an adequate way of identifying countries, each of which get their own ccTLD, there might be other geographic identifiers that would prove useful. For example, cities and towns aren’t well served by the country namespace - is LONDON.CO.UK a better choice than LONDON.COM or LONDON.ORG? London is a major international city whose reach and influence extend far beyond the UK. It may be that none of these domain name suffixes would adequately describe the Mayor’s office, Tourist Board, local authority or Chamber of Commerce. Perhaps LONDON.CITY would be better.

Geographic identifiers need not be restricted to the ISO3166 list. Individual countries are not homogeneous, nor are they even strictly defined. Current developments in places like Kashmir and Northern Ireland may result in places where sovereignty is deliberately vaguely defined. The finality of the ISO3166 would let down the inhabitants of those places.

2. New Types of Entity

Most recent activity within the DNS stakeholder community has been centred on the introduction of new “generic” and “sponsored” TLDs. While most of these new TLDs do follow precedent by addressing types of entity not sufficiently addressed by the current namespaces (.BIZ, .NAME, .AERO, .PRO, .COOP), some have instead been focused on the presumed content of the websites that the domain names would be used for (.INFO, .TRAVEL, .JOBS, and perhaps .XXX sometime in the future). One proposed TLD, .MOBI, is even predicated on the type of client device that would be used to access services on hosts within that domain. This change in the semantics of top-level domains has not been met with universal approval. Tim Berners-Lee has reservations about the .MOBI proposal, arguing that adding content-based addressing rules at the DNS level breaks the most basic principles of equal and unfettered access to services and information without discrimination.

What other kinds of entity aren’t yet serviced by a top-level domain? .KIDS would be a TLD that might meet our criteria, although it too has mostly been marketed on the basis of content.

It seems to me that the case for new generic top-level domains is overstated. There are very few entities that are not already serviced by the existing TLD space. But there are plenty of companies who want to make money running a registry and lots of marketing people who want to make money selling new TLDs. That said, there is also a real need to grow the namespace to meet the demands of Internet users.

Extend “Downwards”, Not Expand “Outwards”?

In the Public Forum, Dr Cerf also said:

“In the architecture of the Internet, the Domain Name System was intended to be a very hierarchical structure so as to avoid extremely large bottlenecks anywhere. You always had the freedom to expand outward. So for many of the functions that one might claim a top-level domain is needed, one could argue that function could do just as well at [the] second-level… It’s an almost fractal-like design.”

Dr Cerf explains how DNS allows for the creation of domain names that add semantics by means of delegation. it is a concept with a strong basis in the founding principles of the DNS - it was Jon Postel who, in 1995, first suggested the idea of creating a domain registry under .UK.COM as a competitor to .CO.UK. There are a number of organisations that provide a “second-level” registry service: as well as my own, NetRegistry provides a registry service under .AU.COM and .JP.COM, and a non-profit organisation provides free .EU.ORG domain names. These domain registries extend the namespace in the same way that TLDs do, but also have the advantage of adding further semantics to the domain name.

I am (perhaps unsurprisingly) convinced that the cultivation of second-level domain name registries within the existing namespace is a Good Thing, and should be broadened to include more generic names, as we have started to do with .WEB.COM. Some input from the DNS stakeholder community, including ICANN, might provide momentum to encourage the registrants of potential “seed” domain names to start offering their own domain registry services, growing the domain name space to meet demand without the lengthy and controversial process of creating new TLDs. And past experience has shown me that when a new competing registry enters the market it often has the effect of increasing the demand for your domain names, not reducing it. Everyone wins.