DNS Turns Twenty


Paul Mockapetris invented DNS while working at the University of Southern California 20 years ago in 1983. It took a number of years to catch on. I still remember editing /etc/hosts by hand on every machine I managed in 1988. Most machines didn't have complete host tables. There were a few complete host tables on various machines spread around the net. This led to a process where sending mail to someone meant that you had to pick the route using UUCP-style ! routes in the email address. The trick was to get the message to a machine that knew about the host you were sending mail to. It was a game and we enjoyed being good at it. DNS spoiled the fun, but made the net much more useful. This ComputerWorld article talks about the history of DNS and its invention. In the article Paul says:

I had expected people to think a little about enhancing the infrastructure, but it's still working hard and I don't think there are any obvious stress points where it's going to fall down in the next couple of years.

It is amazing that it continues to work, but that's a testament, I think to the decentralized, albeit hierarchical, architecture of the the DNS system and the overriding simplicity of what it does. It just maps domain names to IP numbers. Contrast that with UDDI which shares a potentially decentralized architecture, but has a much more complicated goal to return services based on more nebulous criteria than a simple, hierarchical domain name.

The article seems to make the invention of domain names and DNS synonymous and I don't think that's right. I believe the notion of hierarchical domain names existed before the DNS mapping system was invented, but I could be wrong. This predates my Internet experience by 3 or 4 years.