review of DNS

The Domain Name Service is the largest and most important service in the
Internet. The first generation of DNS was using a centralized HOSTS.TXT
file to provide mapping between host name and address. The HOSTS.TXT was
obviously a disadvantage. In early 1980s, distributed databases were used
to design and implement the Internet Domain Name System.

This paper was written in 1988. The authors studied the initial design of
the DNS, listed what they believed to be the surprises, successes, and
shortcomings of the system. They also predicted the future directions of
DNS.

Among the six successes listed in this paper, three of them are related to
the administrative hierarchy of the domain names, including variable depth
hierarchy, organizational structuring of names, and mail address
cooperation. The DNS servers are organized according to the administrative
hierarchy. With the increasing hosts and users in the Internet, new
techniques are proposed to replace the hierarchy, for example, the DDNS.
DDNS uses a peer-to-peer Chord-based distributed hash table to serve DNS
records. DDNS provides the distributed administration for name data with
good load balancing and fault tolerance performance. For the hosts, which
have no fixed IP address, DDNS manages their name data using the
self-organizing and adaptive nature of Chord.

The DNS security problem also should be considered, since IP address can
be forged and thus it is possible for malicious people to impersonate DNS
servers. DNS Security Extension (DNSSEC) developed in late 1990s provided
a mechanism for clients to verify that data they received from the servers
are authentic. In DDNS, Chord guarantees the security to some extent.

Caching technique used in DNS is to speed up the service response time.
However, the update of domain data may cause the information inconstancy
between the cache and current domain name address. The time-to-live field
is to indicate the length of time that the cached data can be reused. It
is expensive to maintain the content of the cache if the TTL is too small,
while it may provides incorrect address information if TTL is too large.
Fortunately, the domain data is not change frequently, a TTL could be
chose based on a tradeoff. DDNS

Although the paper was written 20 years ago, it was thought-provoking at
that time. Many later researches have been done for the shortcomings, and
even the successes.
Received on Sun Oct 02 2005 - 12:43:21 EDT