It gave me a warm, nostalgic feeling when I read of ICANN’s proposed agreements to let domain name registrars raise
prices arbitrarily on domain names. This is what the Domain Name
System is all about–at least, since I got involved in policy debates
around the Domain Name System about 1994: milking a monopoly for all
the cash you can get. But the issue also made me re-examine an old
question: in place of the ordained chaos of the current governance
policy, is it possible to administer the Domain Name System in a way
that apportions names in a rational manner?

The rent-seeking history of DNS

The thrust of the proposed ICANN agreement with three top-level
domains (.BIZ, .INFO, and .ORG) is simple: a registry could charge as
much as it wants for a domain name once it finds out if it’s popular.
Thus, if some scrappy startup-in-a-dorm-room registers
GitYourCheapMovies.biz and it turns into the next Internet sensation,
the registrar could raise its fee to a million dollars.

Admittedly, there’s a built-in lag: the registrar would have to give
six months notice, and the site registering the name could lock in the
old fee for ten years. But with commerce on the Internet becoming
mainstream, this is inadequate protection against exploitation. The
business of registering domain names becomes like the old joke of the
gunman who demonstrated perfect marksmanship–until he admitted that
he shot the holes first and then drew the targets around the holes.

This is, of course, all about ICANN supporting its own bureaucracy, a
quest that has been an almost perpetual crisis since ICANN began. The
fees paid by registrars to ICANN are based partly on what the
registrars charge their clients (see, for instance, paragraphs
7.2(a)(ii) and 7.2(a)(iii) of the .BIZ agreement). So the proposal is just a particularly scary
variation of an old trend: to siphon off from innovators some of the
substantial revenues that the Internet makes possible, and use it to
benefit those with a choke-hold on the indispensable Domain Name
System.

The policy debates around the Domain Name System began in the
mid-1990s when Network Solutions decided to charge $100 to register a
.COM domain, and $50 to maintain it hereafter. Because maintaining the
computers that serve up .COM domains costs a fraction of a penny per
domain name, this was what I called Information Highway robbery, and
it generated (along with various other abusive practices) a revolt
among Internet activists.

The cost of domain names has gone down drastically since then, but
egregious rent-seeking behavior has remained. ICANN created a huge
superstructure that does nothing very much, and does it mostly to
perpetuate ICANN’s own dominance, rather than things that would
actually improve the Internet for its users (such as promoting IPv6
adoption, or measures to secure DNS such as DNSSEC, or truly
internationalized domain names that are fit for every
language). Flirting with bankruptcy every couple years, ICANN has
tried to fund this superstructure through various fiats such as trying
to collect one dollar per year per domain name.

Can we leave this sordid history behind and look toward a future
Domain Name System that offers sensible policies and enhances the
user’s Internet experience?

Three views of DNS

Three general schools of thought address the administration of
DNS. I’ll arbitrarily call them the structuralists, the libertarians, and the indifferent.

The structuralists

The structuralists want the Domain Name System to represent a grand
knowledge domain, a kind of Dewey Decimal system like that used by
U.S. libraries. (Librarians are very interested in knowledge domains,
also called ontologies.)

Anyone could easily come up with an elegant breakdown that is much
more rational than the current odd division of TLDs into .EDU, .GOV,
and so forth. The current ccTLD (country-code) domains make good
sense for certain domains (such as governments) but are increasingly
obsolete for business and NGOs, which often cross national
boundaries. So one could create a .CULTURE domain, for instance, which
was subdivided into FINEART.CULTURE, MUSIC.CULTURE, and so on. The
world of information and life itself could be partitioned by concepts.

One problem with such a system is that everyone who tried to invent
one would come up with something unique, and it would be hard to find
a consensus. Furthermore, the system would have to be incredibly
flexible (a problem that dogs the Dewey Decimal system). For instance,
if an ontological DNS were created in the mid-1990s when the policy
debates began, it would have been hard for anyone to imagine a place
for citizen journalism (except for the handful of “Internet democracy”
visionaries I hung out with). But now the blogosphere has clearly
created one of the most important categories of Internet behavior.

If domains are organized logically, it starts to make sense to
associate policies with certain names. For instance, a domain could be
allocated to aboriginal peoples of the world and could require that
registrants meet certain requirements.

But there’s a downside to specialization. Many proponents of the
structural approach like the idea of a .XXX domain where erotic sites
can be sequestered. Proponents of free speech are very leery of this,
because it can lead to onerous censorship: some authority could
require a controversial site to register under .XXX, where it might be
filtered out by schools, libraries, and ISPs.

A structural approach, therefore, would almost invariably lead to
bureaucracy and fossilization.

The libertarians

The libertarians deal with the problems of structuralism in a
delightfully elegant way: they call for the removal of all limitations
on who can register TLDs. There are currently 13 hosts (root servers)
on the Internet that serve up the official DNS. But numerous other
sites offer alternative root servers, and anyone is free to point his
personal computer at any number of them. Various sites have registered
fanciful TLDs with alternative servers.

There are technical debates over the strain that root servers might
experience if everybody and her sister could register a TLD. But
proponents of the libertarian system point out that the .COM servers
have no trouble serving millions of .COM domains, so root servers
should be just as scalable.

The other major difficulty with the libertarian approach is that
different root servers could choose to register the same TLD for
different sites; there would then be an inconsistency in the Domain
Name System. A popular TLD such as .PORN might point to different
sites depending on which root server your computer is pointed to. Some
people have suggested a super-registar that would adjucate disputes,
but this just reproduces all the problems of ICANN at a higher metalevel. Other
people suggest a first-come-first-serve policy for TLDs.

And every system has to deal with disputes that are external to the
system, such as claims of trademark infringement. In the 1990s, with
courts loaded down by domain name disputes, national legislatures
passing draconian bills, and the World Intellectual Property
Organization proposing restrictive policies friendly to major
corporations, ICANN made a modest attempt to bring disputes in-house
and institute some protections for small businesses and non-profits.
If this system (however flawed) is replaced by something looser, the
whole judicial/legislative/WIPO mess will start up again.

The indifferent

The third major school of thought on Internet domain names says that
they are increasingly irrelevant thanks to search engines. If you can
enter a search term into Google and find what you want within the top
ten hits, what do you care whether it has a catchy domain name? And a
lot of people find sites from links on blogs or trusted portals; they
hardly even notice what the domain name is.

To take this argument to its logical conclusion, we might as well
chuck DNS and just use IP addresses. The sole remaining justifications
for domain names are that sites have to balance the load of traffic
among multiple servers, and that domain names move between servers
faster than IP addresses do.

Because these are good reasons to keep DNS around, it would make sense
to try to find a rational system for divvying up the name space, the
goal at the beginning of this article.

Maximizing choice

There is no doubt that the current system of TLDs is irrational. ICANN
has created sites such as .AERO and .MUSEUM that nobody uses, while
denying requests for perfectly reasonable TLDs and giving in to
censorship demands for an .XXX domain (not to mention overriding the
use of .BIZ by alternative root servers).

We have already gotten used to domains associated with various
policies. Everybody understands that FACEBOOK.COM registers sites for
individual students. It would be natural to take this to another level
and register a meta-domain that set policies for people registering
subdomains. A new competitive market could be created for desirable
domains, such as the idea for aboriginal sites mentioned earlier.

But this doesn’t require changing the current TLD system. A market
could grow up in second-level domains. If it seemed to meet a need
(right now, it’s not clear whether there is a need) a successful
system of second-level and third-level domains could become a model
for reforming the TLD system.

Of course, the ICANN proposal for a domain fee free-for-all casts a
pall over any such innovations. Stability at one level of DNS would
permit innovation at higher levels. But so long as there’s ordained
chaos at the top, Internet innovators have extra risks to consider.