From: Karthik Gopalratnam (karthikg@cs.washington.edu)
Date: Wed Nov 17 2004 - 04:19:16 PST
Review 23 - DNS
This paper presents an overview of the state-of-the-art in DNS in 1988.
It talks about the original design goals, the then implementations and the
various important features of DNS's evolution upto that point. DNS was
proposed as a solution to an internet wide "HOSTS.TXT" file which had to be
present on all the machines in the internet for the nodes to communicate
with one another. The paper presents several considerations that went into
DNS - the hierarchical strcture, "thin implementation", caching without
losing performance guarantees. This resulted in a distributed and scalable
naming system. In particular the original design had many parallels to the
Grapevine system of Birrell et. al.
This paper does a great job of laying out the successes and shortcomings of
the system. The structured variable depth hierarchy was a particularly good
design result - the semantics of the naming service were essentially
unbounded and could be used in versatile ways. The fundamental intelligence
in this design has seen DNS evolve with little change until the present. One
of the biggest flaws was that any discussions of the guarantees on
security/reliability were completely absent. The root servers are quite
susceptible to attack and essentially provide a single point of failure,
forcing the hosts to r
ely on cached copies of names instead.
This archive was generated by hypermail 2.1.6 : Wed Nov 17 2004 - 04:19:19 PST