Features

net.wars: DNS 21 years old enough to drink

Of course it's a PR stunt to lay on a party and a cake to celebrate the 21st birthday of the domain name system.

But it's a good cause: the Net as we know it would probably never have taken off with such speed and impact without it. Most people think that graphical interfaces and the Web were the most important elements of making the Net user-friendly enough for the masses, but I've always inclined to the view that the domain name system was the most important contribution to Net usability. Without it, we'd all be typing IP addresses, and while people can obviously learn strings of numbers - cf the telephone system - they don't like learning too many of them.

Paul Mockapetris, who invented the domain name system and gave it to the world (although of course it's been extended and improved since then) was in London on Wednesday to celebrate the birthday - and promote his company, Nominum. He had some fun numbers. He estimates, for example, that there are 2 million DNS servers on the public Internet. If each costs, on average, $500, that's $1 billion in hardware. On private intranets, at a guess? 10 million servers. Who pays? Someone asked. In the best cooperative tradition of the Net, we each pay for our part, whether it's our server or the bandwidth to our ISP's server.

What's more fun is his reminiscences of the arguments – many of which continue to this day – about how to structure the DNS. Country code domain names – like .uk and .fr – were a popular concept, although there were competing demands for generic top-level domains.

"The only reason we have .com," he says, "is that people got tired of arguing with me about it. They said, ‘It will never catch on, so it won't do much harm.'"

That touches on a continuing argument: should the Net be geographically organised or not? There are times when you want it to be. For example, if you're looking for a place to buy a washing machine, it's convenient to be able to tell from a country code domain name such as .co.uk that chances are the company whose Web site you're browsing will deliver to you. Most of the time, though, geographical divisions are just annoying. I don't want Google to return hits it thinks someone in the UK would want to routine queries just because I have a UK-based IP address. I want the best results, the right ones, the ones I'm looking for, which often won't be in the UK at all.

Questions of geography are just part of the broader question of what the DNS is actually for. It started out as a way of making it easier for users to route email and other Internet connections. Instead of having to know the exact address of the machine you were sending to, you could use a memorable name and the DNS would do the rest. Imagine if you had to know one IP number to send me email and another to log onto my Web site instead of simply remembering the name pelicancrossing.net. This is why I believe the DNS is the fundamental technology that made the Net usable for the mass market. By the time the DNS was ten years old, though, it was clear that people had begun thinking of the DNS as a kind of automatic directory. The expectations of lawyers and sales and marketing personnel was that names should reflect real-world brands, trademarks, product names, and so on. And they were complaining: too many of the good domain names were taken. Now, of course, we see how silly that was. We should have just told them to wait a few years and things would be fine. But instead the various Powers That Were launched into a complicated revamping of the DNS to try to create more "good" names, which brought us ICANN

In 2003, the DNS held roughly a billion names; Mockapetris was predicting the number would double every year for the next five years at least. At that rate, there will be more domain names than people by 2006. The reason Mockapetris expects the DNS to get so much bigger: the desire to make several very large spaces of numbers intercommunicate: the telephone network, IP numbers, product barcodes, the ISBN system for identifying books, and the serial numbers embedded in RFID chips.

In a very real sense, though, as you read about new developments such as foreign-language domain names (actually a Unicode veneer over extraordinarily ugly ASCII strings), you realise that to a very large extent still no one has solved the fundamental question of what the DNS is actually for. Does it follow geography, trademarks and company names, or types of users? Is it a directory or a marketing construct? Should names be automatically guessable? What about, instead, dividing up the Net by language? Or registered company names, like .plc.uk and .ltd.uk? Or content, like .xxx or .kids? Why not have an electronic commerce space where retailers register according to the areas they deliver to? Is one of these more right than another?

Mockapetris is like Switzerland. "There can be more than one train of thought," he says.

So, if there is one question you have to ask Mockapetris, it's this: "If you were doing it over again, knowing what you know now, what would you have done differently?"

He grins, evilly. "I'd have owned it!"

Happy birthday.