From: Rosalia F Tungaraza (rltungar@cs.washington.edu)
Date: Sun Oct 03 2004 - 23:15:54 PDT
This paper provides the history behind DARPA's IP/TCP protocol; the
factors that motivated its invention, underlying architecture, and some of
the modifications that have been performed to attain some of its present features.
Comparing that protocol's architecture to an hourglass with the IP layer
being the neck, the author did a relatively good job in providing the
rational behind the opposite sides of the neck; the transport layer
followed by the applications layer on one side, and the network layer on
the other side of the neck. He elaborates on why and how the design ensured that different types of networks and
transport protocols (e.g. TCP and UDP) could be part of the Internet.
Moreover, the transport protocols were implemented in such a way that they
could also support different types of application programs (services).
Nevertheless, he neglects to expound on the neck region of the hourglass,
namely the IP layer. Being such a crucial region in the architecture, one
would expect to read more about it. For instance, why it was necessary
either from the design or implementation's point of view to separate the
TCP from the IP layer. Hence, one way to improve this paper would be to
include more information about this layer.
The paper/work proposes a number of open problems. One such problem
involves determining the performance of a particular implementation of the
IP/TCP protocol. The author does not claim to have a solution for this and
thus, a possible future work could be geared towards analyzing the
performances of different implementations of that protocol
(realizations of it). Eventually, one could propose some standards that
would maximize the performance of a given realization. Another open problem
suggested in the paper is to explore alternative building blocks (other
than datagrams) for the DARPA Internet protocol.
This archive was generated by hypermail 2.1.6 : Sun Oct 03 2004 - 23:15:55 PDT