From: Susumu Harada (harada@cs.washington.edu)
Date: Sun Oct 10 2004 - 12:12:50 PDT
"A Protocol for Packet Network Intercommunication"
Vinton G. Cerf and Robert E. Kahn
This paper provides a detailed description of the implementation of the
initial version of the Transmission Control "Program" to enable processes
on different hosts connected over a packet switched network to communicate
with each other. The paper presents issues that are involved in
attempting to connect multiple packet switched networks, basic principles
which should be followed in order to make the interconnection work, and
specific implementation guidelines for how to go about enabling such
interconnection.
I was surprised to see the level of detail this paper goes into in
outlining the implementation details of the TCP protocol. This is in
stark contrast to the previous paper on "Tussle in Cyberspace", which
didn't give any details on how the presented problem could be solved.
It was interesting to note how the authors referred to TCP as Transmission
Control "Program", and not "Protocol". This seems to suggest the
underlying notion that at the time the paper was written, TCP was one
monolithic "protocol" that was attempting to accomplish the task of both
(what we now call today) TCP and IP.
The paper does a great job of detailing the specific characteristics that
their proposed network interface protocol should possess, such as
separation of protocols at the inter-host level and inter-interface
level, layering of protocol abstraction to provide encapsulation of
transport functionalities, addressing mechanisms, and packet segmentation
and reassembly.
The paper also not only presents a single solution to each aspects of the
problem, but addresses several alternatives and discusses their strengths
and weaknesses. For example, in the section on process level
communication, it considers two scenarios in which the source TCP attempts
to send multiple segments destined for the same TCP but different
processes.
There are several issues that the paper seems to have glossed over. One
is that of security, where the only mention for addressing the issue is to
have some sort of pseudorandom port address switching agreed to by the
sending and receiving processes. Apparently, back in their days the issue
of malicious hackers and eavesdroppers were not of primal concern.
Another point that could have improved the paper is the inclusion of some
form of empirical results from simulation or experiments. Of course they
mention in their conclusion that such next step is necessary, but at least
some basic test results demonstrating the soundness of their proposed
methods (e.g. retransmission algorithm) would have been nice.
Overall this paper provides a relevant historical context of the evolution
of the TCP protocol, and serves as a useful reference in noting how the
protocol has evolved over the past several decades and in how the current
Internet protocols might evolve as our use and extend of the Internet
grows.
This archive was generated by hypermail 2.1.6 : Sun Oct 10 2004 - 12:12:50 PDT