From: Ethan Katz-Bassett (ethan@cs.washington.edu)
Date: Wed Nov 17 2004 - 07:49:21 PST
This paper presents an overview of DNS, some of the design decisions that
informed it, and information on its early use. It was an interesting read,
as I use DNS every day, but have never really thought much about what is
actually happening. Before DNS, HOSTS.TXT was used to map from names to
addresses. The file was centrally maintained and then sent to all hosts in
the Internet. Obviously the system was not going to scale well; with the
spread of the Internet today, that system seems positively antiquated.
According to the paper, HOSTS.TXT was in trouble once users began to connect
from individual workstations. At Berkeley, according to the data provided,
there was a 10 scale growth in just 2 years.
DNS, in contrast, is quite flexible and scalable. Probably most
importantly, it gives a distributed system for handling the names and
addresses. As a user, the variable-depth tree based naming seems like a
sensible choice. I had not realized that, even at the early date when the
paper was written, the top level domains EDU and COM were already in use.
The authors note that it was difficult to make performance measurements;
were they feasible, I would have liked to see data on how DNS performed in
these early years. They discuss how many negative DNS responses they saw; I
am curious what the percentage is now and how much negative caching is used
as a solution. I imagine some misspellings come up quite frequently.
This archive was generated by hypermail 2.1.6 : Wed Nov 17 2004 - 07:49:26 PST