ICANN is Not a World Government

Interview with Vint Cerf, ICANN's newly elected Chairman of the board, who wants that the non-profit company to focus on its challenging technical tasks

It's hard to remember all the institutional affiliations and board memberships that Vint Cerf serves. As one of the fathers of the Internet -- he co-invented TCP/IP with Bob Kahn back in 1973/74 at UCLA in a project for ARPA, the Advanced Research Project Agency of the U.S. Department of Defense -- he was a founding member of the Internet Society, he is engaged in the Internet Steering Task Force as well as the Internet Engineering Task Force, and at the last annual meeting of the Internet Corporation for Assigned Names and Numbers (ICANN) in Marina del Rey he was elected Chairman of the board after being involved with ICANN from its very beginning in autumn 1998. There was just one vote against him - by Andy Müller-Maguhn, the newly seated European At Large Director.

In der Nähe der ICANN-Tagung in Marina del Rey

Müller-Maguhn dismisses Cerf mainly because of his (Cerf's) industrial background: the new Chairman is actually Senior Vice President Internet Architecture & Technology for the telecom giant WorldCom and also a big time lobbyist at the telco and media company driven GIP, the Global Internet Project, which calls in its mission statement upon governments "to encourage private sector solutions to Internet policy challenges." But Cerf defended his role recently in an Email to Hans Klein who put him in the telco "supplier" pigeon hole in one of his newsletters.

"In the GIP, I do try to represent an industry perspective -- that was the intention in forming GIP," Cerf said. And he continued: "However at ICANN, I try very hard NOT to represent industry alone or the protocol/technical community, alone, or any other group alone. I see the job of ICANN director as much broader than that and that ICANN, and those who depend on its choices, are not well served by narrowly viewed perspectives."

Stefan Krempl sat down with the 57 year-old networker in his suite in the Marina del Rey Beach Marriot after the seven hour debate about the introduction of new generic top-level domains (gTLDs) to discuss the main tasks he sees for his year-long term as the Chairman of the board, the role of governments within ICANN, the meaning of the new domains as well as the notion of ICANN as some kind of a "world government".

You've been fostering the development of the Internet for so long now. How far do you think did we come in building a true international medium?

Vint Cerf: We still have some distance to go. The Internet is available in an enormous number of countries, something over 200 countries around the world. However, it's penetration is very limited in some places. And plainly, there's a lot of interest in having character sets others than Roman as part of the domain name system in addition to be able to use non-Roman characters in the World Wide Web, which has been a big help in terms of internationalizing the Web already. There's is quite a difference, though, between putting those character sets into ordinary text on a web page where there is little or no interpretation required and having characters and strings that programs are going to use to do something with -- to compare things with other things or words that have meaning because now people are relying on a domain name to mean something, if it's a product name or a trademark or so on. And so we've uncovered good soil, but also many unknown problems in introducing the new character sets into the domain name system.

I've been quite surprised to understand how very dependent we have been on a simple subset of character strings. We've been using IA4 (International Alphabet 4), that doesn't allow very many characters. It's the alphabet and the numbers zero through nine and a hyphen -- that's all. That particular set doesn't know any difference between upper case and lower case, what has become very important with email where you don't want to care about such things.

So, domain names momentarily have such a nice character because of the simple set that they use. And that has made it possible to drop domain names in almost every application without any difficulty. But when we start putting in very difficult character sets or encoding 16 bit character sets in 8 bit ASCII, it's much more messy. And so, making the applications work when we introduce that level of richness is much, much harder than we thought. Only in the last two or three months I found out how convenient the very limited domain name character set was and how much it helped us. So we got a lot of work to do there.

Plainly, the Net has not spread uniformly around the world. But I am impressed by the statistics that less of half of the Net usage is in North America now. The other users are pretty widely distributed. We saw big jumps in Europe where in summer more than 90 million people out of 300 million people were online. And I also think we're only just beginning to understand what the social and economical aspects of the Internet are.

There was quite a silly economic effect for quite a while now on the stock market. I think that particular phenomenon has begun to dissipate, in some cases in a very dramatic way. But as much fun as it is to watch that kind of craziness, I am very relieved to see that the Wall Street analysts are starting to pay more attention to the business plans of companies that are working in the Internet space. It's reassuring to know that at least some believe it's good to make more money than you spend.

After the dot-com phenomenon seems to fade, will we see a rush towards the seven new domains or at least towards dot-biz now?

Vint Cerf: Frankly, I don't know. One of the reasons why we wanted to open the top-level domain space in a relatively measured way was to try to gage what kind of a rush there is. My expectation is that many people who have registered in other, similar gTLDs will want to register in dot-info or dot-biz for example. I don't know what would happen if we would have a very large number of new TLDs. I do know that we have not seen a rush simultaneously in the ccTLDs and in the gTLDs. Many companies that have registered in dot-com have not registered in dot-us or dot-fr or dot-de. If you have a physical operation in Germany, for example, then registering something in dot-de makes sense if you want to call attention to your German presence. But if you want to direct people to your global, multinational presence then having something registered in a gTLD is far better. Though, it's very hard to judge. The faint and smart people might feel like having to register in everything, but other people may not.

There were also other proposals on the list like dot-sucks. Why didn't they get the vote?

Vint Cerf: We talked about the Name.Space proposal which did have quite a variety of names including dot-sucks. They had a business model that was not completely focussed, however, on free speech or something like that. They actually very focussed on advertising. To my understanding they had quite a large number of new domains because they thought about categorizing either products or services or activities. Their hope, of course, was, as I read it, to persuade people that they should put advertising in various of these top-level domains, that they would advertise something like dot-sports. They were not approved in this round, but after we have some experience with this first set -- which is pretty diverse -- one hopes to go back and reexamine the other proposals.

There have been a lot of trademark wars in dot-com and other TLDs. Do you expect to see all these problems showing up in the new gTLDs, too?

Vint Cerf: For the ones that are more generic and open like dot-info and dot-biz I would expect the same issues to arise there, that have arisen elsewhere. I don't think we really entirely come to grips with the collision of the domain name space, which requires uniqueness, and the trademark space which permits multiple users. And even if you separate things into different top-level domains -- it doesn't work out.

I used to work for the telephone company MCI before we departed to WorldCom. And there's also a bus company in the Midwest that is called MCI. So we could figure out to register www.mci.com. And the bus company could take dot-biz. But dot-biz or dot-com doesn't give you any clue about the company behind it. There are many categories that you can register in the trademark world, but I don't know if, even if we made them all top-level domains or second level domains, if this would carry any intuition for people. This "class 16", what would this mean to anybody. So, it would have to be something else, something in the string that would reflect a special class. But then we would get to the question what to do with different classes in different trademark offices around the world.

No one has really figured out a way around this collision. One thought that crossed my mind was to say, well, what would happen if you couldn't make a claim to trademark preference or priority just because you had a trademark in some class. Supposed you would have to go to the trademark office and register your trademark in a new class called domain name, it still would be the case that only one person would get the register in this class. But at least you would have gone on the record as having registered for that particular class as opposed to the class of advertising or association membership or something else.

In a way this would shift the burden of collision to the trademark office rather than on ICANN, which I always think is a good idea. So if you went to the trademark office and registered your trademark specifically in the class domain name then within the scope of that trademark office authority you would have priority for that domain name. Now, some people would argue, that's not fair, because you can have two trademarks for ABC, one for skies and one for cars. And when we get on the Net, only one of you can have ABC.com, even if both of you registered your trademark legitimately in your class. The only way to get out of this is to say that neither of you has priority to the domain unless you have a trademark in the domain name category of trademarks.

This sounds good on the surface, but it might be wrong. The reason is that if you and I both successfully get trademarks in different business, we both want to be able to go to the newspapers and magazines and publish our trademark. If you think of the Internet as a publication service then we aren't allowed having two people with the same trademark. That's treating domain names as advertising strings -- and I am not happy about that. But that's what people are doing. If you had to register under a class called domain name, at least one person could win. Another interesting argument: if you're given a telephone number in the U.S. you can look at that number an you can try to find some letters on the touch-tone pads that use to spell out a word and each letter of course maps to a specific number. If you find a word you like and I find a word that I like that would map to the same number, only one of us wins. Even if the strings are different, you can't have two guys with the same telephone number. And everybody accepts that. So that's a little bit like not having two people with the same domain name.

Would it be a solution to have telephone numbers in the higher parts of the domains? There were a couple of proposals suggesting something like that.

Vint Cerf: We didn't want to go down there to authorize that yet. And the reason is that in the event that they are intended to refer to the telephone network that it's really important to make sure that an individual or an entity that has been assigned a telephone number doesn't have that number hijacked away from you by some other party who registers that number and makes it point somewhere else in the Internet. So, if you were a registrar and I somehow walked up to you and said, "register this number,Ó and it happens to be your telephone number and they don't check, then I'm taking your telephone number. And if people use that number, they go to my site. That can't be what you have in mind.

There are some customer protection issues here. And I think what the ITU understandably was saying is that they have the responsibility to assigning those numbers. And so any mechanism that has a third party coming up and making assertions about a telephone number he claims are his there has a way to double check that. The best party to do this would be probably the PTT. Then we would have this hierarchical responsibility on the numbers based on which PTT got it. I suggested to the companies that made proposals in that area that they'd better look very seriously at this use of getting hold of a telephone number in the domain name space and having a way to verify who has access to these numbers.

There's also been quite a lively discussion about content related gTLDs like dot-kids. That brings us to the point that many people think about ICANN as some sort of a world government.

Vint Cerf: Aaaargh.

What? Can't you hear it any more?

Vint Cerf: This is not a good way to think about ICANN. Many people want it to be. I think I understand what's going on. People see this Internet, and they see a lot of its power and effects. And they recognize, it can be abused. And they want somehow to have a place to go to deal with problems and their different policies. Now they see ICANN over here and it has a fairly international profile. So they want to load up ICANN with anything that might have to do with either regulation of the Net or with somehow addressing complaints. This is probably not a very fruitful idea in my opinion.

ICANN will function best if it's very constrained in its responsibility. This does not mean that the board or the staff are insensitive to some of the issues raised, but rather that ICANN is not the right place to resolve these issues. So I would much rather prefer to see these matters raised in forums where those issues are commonly handled so that ICANN can stay pretty focussed on things like: does the registry work properly? Does it interwork with the root? Does it interwork with clients, registrars? Do you have a way of backing it up in case something goes wrong? There are pretty straight forward technical requirements that you have to meet. And to meet those we shouldn't have to assess so much more, like dig into our funding for example.

There's some cooperation between ICANN and the governments of the world in the GAC, the government advisory council. People sometimes are suspicious what's going on between the two sides.

Vint Cerf: Absolutely.

Could you please elaborate on the role governments have within ICANN or what role you want them to play in the future?

Vint Cerf: Well, at the moment the GAC has no role in the selection of the board members. That might conceivable change. I could easily understand a government saying: we have responsibility for a top-level domain and the decisions made by ICANN could have an effect on our top-level domain, an important countrywide infrastructure. And so they want some representation. So I can imagine the GAC becoming more like a support organization. GAC can also be a very good avenue for discussing various issues associated with Internet operation, because government and the private sector do have to work together in this field. They may disagree with each other. That's not unusual. But at least there would be a forum for having a debate.

Right now it's a little less structured than that. A top-level domain is typically run out of the government's authority, often it is run by a university, but more often it is run by a government sanctioned operation which might or might not be for profit.

Parts of the basic Root Server are still under the control of the U.S. government. Do you want this to change?

Vint Cerf: Well, the A-Root Server is operated by Verisign Global Registry, but any changes that are made to the top-level of the system, to the "dot" of the system, have to be approved by the Department of Commerce. There are several requirements that we have to meet before the complete transfer of responsibility goes to ICANN. One of them is to get to the point where we have essential agreements with the various ccTLD operators. This leaves open the question if we should also have some kind of formalized relationship with the country government that is associated with the ccTLD. We're still kind of wrestling with that problem.

I'm naive about this and think it would be a good idea to have a more formalized relationship, so if there's a problem you would have a place to go. Other people are not quite as sanguine about that. Sometimes it's a question of experience. If you're government is mostly, you know, OK, or legal and doesn't take advantage of anyone, that's one thing. I on the other hand you regularly have to do something to defend against some form of governmental abuse, that's a whole different story. Then you don't feel really comfortable with the idea that your government has the central authority over you. So we get both, complaints from governments and complaints from individuals about their governments.

We've seen the first At Large elections in October which got quite a lot of attention, at least in Europe.

Vint Cerf: Oh yes, they did. Especially in Germany, if I recall it correctly, where Der Spiegel was quite vocal about everybody sign up and vote. What, by the way, feeds the notion that ICANN is somehow a world wide government organization. I regretted the impression that that left, because first it helped to overhaul the system which made the whole process a lot less effective and second it somehow feeds the idea that ICANN somehow should take on huge responsibilities that I don't think it can manage.

What role do you see for the At Large community now?

Vint Cerf: I think we have to have a way for communicating with people who use the Net. It is really important that we hear what they have to say. So I can't imagine that we should suspend the At Large list. What isn't so clear is whether we need to manage that group of people as if it were part of the ICANN structure like another support organization or whether it is a more informal source of interaction. It's just so big compared to all the other support organizations that it could consume a lot of overhead just to manage it alone. We have to wait what the study says that's being launched soon to look at the whole process.

Do you think that the so-called "boardsquattersÓ -- the four interim directors still in place -- should resign and hand over their positions to the At Large community?

Vint Cerf: Boardsquatters?

Well, didn't you see quite a lot of guys with buttons at the annual meeting demonstrating against the "boardsquatting"?

Vint Cerf: Oh, I must have missed that one. But now I know what you mean. In Cairo we had a lot of debate about the method of how to integrate the At Large members. In the course of this sometimes noisy debate a compromise was reached not to elect nine, but five of the At Large directors by direct elections as opposed to an indirect process. And at the same time the agreement was made that only five new board members would be elected in this first go. Then there was a debate whether we should keep the other four seats open. And there was some discussion about what the bylaws said about that point and I came away thinking that if we didn't keep someone in that position somehow we might have shrunk the size of the At Large group. So I took the view -- which I think was relatively well received -- that we needed to make sure that there was no question in anyone's mind that there were nine At Large positions, not five. So leaving four people in place was a way to make sure that these slots didn't disappear.

What do you see as your main tasks as the new Chairman of the board?

Vint Cerf: There are actually several. One of them is to do everything that I can to help the organization reduce its scope of responsibility so that it stays more narrow focused on the technical standards that are needed. I don't mean that it should develop the standards, but rather that it uses the standards as a way of designing whether someone is capable of running a registry, running a registrar, what the policies are for address allocation, what policies there are for the domain name allocations. Those things almost need to end up being technical criteria. And as long as you can meet them, it should be OK to enter the operation of a TLD.

We aren't there, obviously. And I think it may take some time before we get all the bits and pieces of the technical framework in place. So my task as I see it at least in the early stages is try to get that process moving on so that we don't have to do what we call "due diligenceÓ on every proposal, where we have to examine its finances, its staffing, its marketing plan, examine its business assumptions. In most other competitive aspects of the Internet you don't have to do that. You just let them go and if they succeed, fine. If they don't that's OK, too, and we get the address space back.

Well, domain names are not quite that simple. And so we have more work to do here.

Vint Cerf: The seven new gTLDs chosen are assumed to be just a first step in opening up the domain name system. When will we see a thousand gTLDs bloom?

Technically, that should not be a large problem. But first we have to get through this initial batch. We have to measure what the side effect is, how many gTLDs are actually in use, what kind of traffic loads do we see on the name server or on the domain name resolver. And then try to apply what we've learned. I hate to guess at numbers, but we'll be mainly through the year 2001 before we see any significant increase in domain names, because I think that it will take that long to have the experience and to come to a conclusion how this all works.

You're also a big fan of the interplanetary Internet.

Vint Cerf: Absolutely. That's why I thought I would steal a little time of the board meeting to announce the successful launch of a satellite carrying the first interplanetary Internet gateway into space, a Sparc engine that's running the interplanetary Internet protocols (Das Internet im Weltraum).

So when will people be able to send and receive information from vehicles in space running the interplanetary Internet protocols?

Vint Cerf: The short answer is today, because there is something running. But it's only in earth orbit. At the end of next year a commercial company hopes to have a new Lunar lander that will land near Apollo 17. It will carry the interplanetary protocols. And then in 2003 we'll launch the first Mars orbiter that will eventually land on the planet with two 150 kilogram rovers seeking water and other forms of life. They will be much bigger than the one we saw a few years ago. And if they find any of that, that would be very exciting. I'm just preparing the research side of this enterprise and I hope to ensure that there is an information downlink from various space platforms (Internetpionier Vint Cerf träumt vom interplanetaren Netzwerk). (Stefan Krempl)