From: Kevin Wampler (wampler@cs.washington.edu)
Date: Wed Nov 17 2004 - 00:01:52 PST
This paper discusses the rationale behind the design of the DNS system, as
well as a brief discussion of the system which it replaced. Given the
nature of this system, HOSTS.TXT, the reasons for the change to DNS are
clear, it required a great deal of manual work, and scaled extremely
poorly.
The DNS system addresses these shortcomings of HOSTS.TXT by creating a
general system for establishing and maintaining name-key mappings in a
distributed manner. It is interesting to see the DNS was not in fact
intended to be used solely for the mapping of domain names. Although this
generality can be seen as a useful aspect, as I am unaware of any other
way in which DNS is currently used it seems likely that a system more
tailored to the Internet, for example in the inclusion of implicit
semantics in a domain name (although this is certainly an arguable point).
I also think that the combination of centrality and distributivity of the
DNS system is reasonably well balanced, as it allows for growth while
staying central enough to keep namings consistent in the different parts
of the Internet. Similarly, the use of the TTL field in the caches seems
to be an effective means of maintaining consistency while allowing the
benefits of a distributed approach.
This archive was generated by hypermail 2.1.6 : Wed Nov 17 2004 - 00:01:52 PST