In an interview with ComputerWire at ICANN’s meeting in Mar Del Plata, Argentina, last week, Cerf talked about the successes, and his concerns, from the four and a half years since ICANN began adding new domains to the internet’s root.

I’m really glad we had a limited introduction, he said, referring to the seven test-bed new top-level domains (TLDs), including .info and .biz, that were approved in November 2000. We ran into a lot of very interesting problems.

ICANN manages the domain name system, at a high level. No new strings can be added to the DNS root system without its approval. But figuring out fair ways to add new TLDs, essentially enabling new businesses, has been one of its major challenges.

Cerfs pointed to the so-called sunrise periods that the 2000 batch of TLD operators implemented in order to help companies register their trademarks before the domains became generally available, avoiding the cybersquatting that had plagued .com.

Registries such as NeuLevel and Afilias, which used sunrise periods in their respective launches of the generic TLDs .biz and .info, saw their sunrise restrictions exploited by speculators that had found loopholes.

A few years after these gTLDs went live ICANN does not have a firm policy on how to approve the next batch. Instead, a year ago it created a definition of sponsored TLDs, gated communities serving specific interest groups, and opened up for applications.

It received ten sTLD applications, and the ICANN board, which Cerf chairs, on Friday unanimously approved the first two of them, .jobs and .travel, to serve the human resources and travel industries.

While it took close to nine hours of strenuous debate for the board to pick the seven 2000 TLDs, it took barely twenty minutes for it to sign off on .jobs and .travel last week.

These are the first TLDs, other than those representing country codes, to be approved since 2000. Three more, .mobi for mobile phone services, .asia for Asia, and .xxx for adult sites, are rumored to be approved very soon.

But Cerf said during a public meeting last week that the notion of a sponsored top level domain may not be well thought-through. He later told us: I continue to be a little nervous about over-differentiation among these different types of TLDs.

We have invented these concepts that were not part of the original design,, he said. It seems to me overly differentiating among various types of TLDs may not necessarily be a good thing.

Part of this thinking comes from recent experiences with the .pro domain, which was approved in 2000 as a restricted TLD. While .pro has no sponsor as such under the ICANN definitions, which would require some level of policy-making delegation to the sponsor, it was intended to be a gated community, much like an sTLD.

While .pro was set up to serve credentialed professionals such as lawyers and doctors, one domain registrar is currently exploiting an apparent contractual loophole to let anybody register any name they want [see separate story].

I’ve come to appreciate that this business has more nuance in it than this engineer realized, Cerf, one of the creators of TCP/IP, said. I’ve also come to appreciate the ingenuity and inventiveness of the people involved in this industry.

The .pro situation came about largely because, eight months after its July 2004 launch, it was flopping terribly. The problem with restricting your customer base is that you get fewer customers, making the business less viable.

Some, I think, have been less successful that the applicants hoped, Cerf said. But I think that as we introduce more TLDs and there becomes more awareness that there are other TLDs and other reasons to register… I think we will see more uptake.

At the same time, the question of whether the internet needs any more TLDs goes still largely unanswered. New TLDs create more cool-sounding available namespace, but some companies feel obliged to spend even more cash to protect their trademarks.

TLDs are not needed to introduce new functionality, but they are often needed, or at least desired, for their brand value. As one sTLD applicant told us last year, it’s a question of architecture versus marketecture.

The DNS is fractal in nature. Anything you can do at the top level you can do at the second level or the third level, and so on. That’s the engineering perspective, Cerf said. The HR sector could be served by jobs.com, rather than .jobs, for example.

But people behave differently depending on which level a name is used, Cerf said. For example, the intellectual property community is much more concerned when something happens at the second level.

There are other tricky balancing acts ICANN has to deal with. On the one hand, it has the stability of the internet foremost in its mind, and a registry going out of business would be detrimental to users who have registered in that namespace.

On the other hand, ICANN takes its contracts seriously, and when a company starts acting outside the spirit of the .pro agreements, for example, ICANN has to figure out what, if anything, can be done about it.

One challenge would be how to effectively assess and induce compliance with the ICANN agreements, Cerf said. The community itself is going to have to help sort through that process.

In addition, ICANN needs some more refined enforcement choices than the nuclear option of revoking accreditation to those that breach contract, to protect those registrants who would lose out if their registrar went out of business, Cerf said.

ICANN will also pay attention to ways to prevent domain buyers exploiting the system. If registrants do things deleterious to the use of the internet, phishing and pharming for example, that can’t be good for the internet, Cerf said.

Recently, for example, a phishing attack emerged that exploits the way internationalized domain names are encoded to fool users into visiting sites where their personal information could be stolen. IDNs should therefore be carefully considered before broad rollout, Cerf said.

I work by the ‘do no harm’ principle. We have to make sure the end-to-end interoperability of the internet is not impaired, he said. He added: There isn’t any better place to solve this. This is our problem.