From: T Scott Saponas (ssaponas@cs.washington.edu)
Date: Tue Nov 16 2004 - 22:41:27 PST
Review by T. Scott Saponas
“Development of the Domain Name System” describes the original design of
the DNS system. The authors also present the reasoning behind the
various design decisions and report some early successes and failures of
different elements of the system.
The main contribution of this paper is providing insight into the design
of DNS. For example the authors explain that it was decided to use
datagrams for make DNS queries because any one request is small enough
to fit in one packet, so if that packet was lost the client could just
send another request. Thus even in the worst case it was as effective
as using TCP. The better parts of DNS are that it distributes the
management of names to the organizations using those names and that it
does this in a hierarchical manner.
There seem to be many drawbacks to some of the design decisions made for
DNS. For example, it seems little thought was put into the format of
the names. While the authors say configuration files will store this
information in the form of labels separated by dots, but applications
can do whatever they want – the authors fail to realize people will
probably adopt one way of representing this information. So now the end
user is actually exposed, for better or worse, to the actual way in
which the names are stored. Also, there does not seem to be much
discussion about who gets what DNS name. Currently, there are multiple
registrars and people have to pay an annual fee to map a name to an IP
(this seems a little weird). This paper basically passes the buck on
this by saying “DNS is flexible enough to accommodate almost any
political choice.”
While I’m a little pessimistic about how names are doled out (possible
because of an ongoing argument with my registrar), certainly the idea of
having a distributed hierarchical naming scheme for address in the
Internet has been successful. One might argue that much of the
commercial success of the Internet has been driven by companies’ ability
to make a brand out of a DNS name.
This archive was generated by hypermail 2.1.6 : Tue Nov 16 2004 - 22:41:37 PST