Clay Shirky's Writings About the Internet
   Economics & Culture, Media & Community, Open Source  

Domain Names: Memorable, Global, Non-political?

May 6, 2002 Everyone understands that something happened to the domain name system in the mid-90s to turn it into a political minefield, with domain name squatters and trademark lawsuits and all the rest of it. It's tempting to believe that if we could identify that something and reverse it, we could return to the relatively placid days prior to ICANN.

Unfortunately, what made domain names contentious was simply that the internet became important, and there's no putting the genie back in that bottle. The legal issues involved actually predate not only ICANN but the DNS itself, going back to the mid-70s and the earliest decision to create memorable aliases for unmemorable IP addresses. Once the original host name system was in place -- instead of -- the system was potentially subject to trademark litigation. The legal issues were thus implicit in the DNS from the day it launched; it just took a decade or so for anyone to care enough to hire a lawyer.

There is no easy way to undo this. The fact that ICANN is a political body is not their fault (though the kind of political institution it has become is their fault.) Memorable names create trademark issues. Global namespace requires global oversight. Names that are both memorable and globally unique will therefore require global political oversight. As long as we want names we can remember, and which work unambiguously anywhere in the world, someone somewhere will have to handle the issues that ICANN currently handles.

Safety in Numbers

One reaction to the inevitable legal trouble with memorable names is simply to do away with memorable names. In this scenario, ICANN would only be responsible for assigning handles, unique IDs devoid of any real meaning. (The most articulate of these proposals is Bob Frankston's "Safe Haven" approach.) []

In practice, this would mean giving a web site a meaningless but unique numerical address. Like a domain name today, it would be globally unambiguous, but unlike today's domain names, such an address would not be memorable, as people are bad at remembering numbers, and terrible at remembering long numbers.

Though this is a good way to produce URLs free from trademark, we don't need a new domain to do this. Anyone can register unmemorable numeric URLs today -- whois says, for example, is currently available. Since this is already possible, such a system wouldn't free us from trademark issues, because whenever systems with numerical addresses grow popular (e.g. Compuserve or ICQ), users demand memorable aliases, to avoid dealing with horrible addresses like Likewise, the DNS was designed to manage memorable names, not merely unique handles, and creating a set of non-memorable handles simply moves the issue of memorable names to a different part of the system. It doesn't make the issue go away.

Embrace Ambiguity

Another set of proposals would do away with globally unique aspect of domain names. Instead of awarding a single firm the coveted .com address, a search for ACME would yield several different matches, which the user would then pick from. This is analogous to a Google search on ACME, but one where none of the matches had a memorable address of their own.

The ambiguity in such a system would make it impossible to automate business-to-business connections using the names of the businesses themselves. These addresses would also fail the 'side of the bus' test, where a user seeing a simple address like on a bus or a business card (or hearing it over the phone or the radio) could go to a browser and type it in. Instead, there would be a market for third-parties who resolve name->address mappings.

The rise of peer-to-peer networks has given us a test-bed for market-allocated namespaces, and the news isn't good. Despite the obvious value in having a single interoperable system for instant messaging, to take one example, we don't have interoperability because AOL is (unsurprisingly) unwilling to abandon the value in owning the majority of those addresses. The winner in a post-DNS market would potentially have even more control and less accountability than ICANN does today.

Names as a Public Good

The two best theories of network value we have -- Metcalfe's law for point-to-point networks and Reed's law for group-forming networks -- both rely on optionality, the possibility actually creating any of the untold potential connections that might exist on large networks. Valuable networks allow nodes to connect to one another without significant transaction costs.

Otherwise identical networks will thus have very different values for their users, depending on how easy or hard it is to form connections. In this theory, the worst damage spam does is not in wasting individual user's time, but in making users skeptical of all mail from unknown sources, thus greatly reducing the possibility of unlikely connections. (What if you got real mail from Nigeria?)

Likewise, a system that provides a global namespace, managed as a public good, will create enormous value in a network, because it will lower the transaction costs of establishing a connection or group globally. It will also aid innovation by allowing new applications to bootstrap into an existing namespace without needing explicit coordination or permission. Despite its flaws, and despite ICANN's deteriorating stewardship, this is what the DNS currently does.

Names Are Inevitable

We make sense of the world by naming things. Faced with any sort of numerical complexity, humans require tools for oversimplifying, and names are one of the best oversimplifications we have. We have only recently created systems that require global namespaces (ship registries, telephone numbers) so we're not very good at it yet. In most of those cases, we have used existing national entities to guarantee uniqueness -- we get globally unique phone numbers if we have nationally unique phone numbers and globally unique country codes.

The DNS, and the internet itself, have broken this 'National Partition' solution because they derive so much of their value from being so effortlessly global. There are still serious technical issues with the DNS, such as the need for domain names in non-English character sets, as well as serious political issues, like the need for hundreds if not thousands of new top-level domains. However, it would be hard to overstate the value created by memorable and globally unique domain names, names that are accessible to any application without requiring advance coordination, and which lower the transaction costs for making connections.

There are no pure engineering solutions here, because this is not a pure engineering problem. Human interest in names is a deeply wired characteristic, and it creates political and legal issues because names are genuinely important. In the 4 years since its founding, ICANN has moved from being merely unaccountable to being actively anti-democratic, but as reforming or replacing ICANN becomes an urgent problem, we need to face the dilemma implicit in namespaces generally: Memorable, Global, Non-political -- pick two.

Subscribe to, the mail distribution list on Networks, Economics and Culture.

To subscribe, send mail to with 'subscribe' in the subject line, or go to