From: Chandrika Jayant (cjayant@cs.washington.edu)
Date: Sun Oct 03 2004 - 23:56:41 PDT
"The Design Philosophy of the DARPA Internet Protocols"
Written by David D. Clark
Reviewed by Chandrika Jayant
This paper explores the early motivations which led to the particular
structure of the Internet and its protocols. The paper is clearly
outlined into a solid introduction, a list of the goals and priorities
of the Internet at its conception, how those goals were met, the
architecture and implementation of those goals, and a more in-depth
reasoning on the decision to use datagrams and TCP. There is adequate
background information in most areas discussed, and the language is
simple enough to be easily understood by those not familiar with
networks. Clark stresses the importance of implementation (i.e. in a
military or civil context) versus architectural structure. He also
prefaces the paper by saying he is presenting "one view" about early
Internet objectives which is important for the reader, in order to fully
realize that there are different potential viewpoints.
There are a few gaps in the paper. For example, Clark
mentions that a multimedia network, though possibly a better solution,
did not fit with the existing structures at the time. It would be
beneficial to the reader to know how much work it would have been to
start from scratch and if the future advantages could have been worth
that cost. Clark should also speak more about hardware ramifications and
their effect on Internet protocols and implementation. Convenience seems
too easy of an explanation to why packet switching was used instead of
circuit switching, as well as why store and forward switching was used.
Finally, some more discussion on the latter four priorities for the
Internet would have rounded off the paper better. It was hard to tell
whether those lower priority goals were hard to meet or if there was
just not much effort to meet them in the military context.
To improve this work, Clark could have given more background
on ARPANET and its motivations. If the Internet had not been created in
a military atmosphere, what would have been different about its
implementation? Did DARPA not think of the future context of the
Internet, or was it always supposed to be just a part of the military
and not a civilian network? These questions could be answered as well as
clarification on the points mentioned in the last paragraph.
Though written in 1988, this paper is extremely relevant
today. It provides an understanding of motivations of one of the most
important evolving technologies of our time, which is necessary for
current design extensions, to see how the Internet was built and what
those protocols have meant. The paper lets us see the importance of
keeping future use/ hardware improvements/ types of use in mind, and how
seemingly appropriate decisions at the time when flexibility is
available will have long lasting ramifications that might be very
difficult or inconvenient to change later on.
Future work suggested is to consider using both packets and
bytes in Internet flow, to work more on Internet error detection, and to
explore distributed management issues. Another suggestion is to
consider using another building block for the Internet instead of the
datagram, which is good for survivability but weaker for the lower
priority DARPA goals. Clark suggests more research on implementing
packet flow accountability. He also suggests a more formal guide for the
implementer of an Internet realization.
This archive was generated by hypermail 2.1.6 : Sun Oct 03 2004 - 23:57:27 PDT