design philosophy of internet protocols

From: Kiran Kumar Gollu <kkgollu_at_cs.toronto.edu>
Date: Thu, 14 Sep 2006 11:12:12 -0400

This paper digs into to present design goals of serveral Internet
protocols
and also provides insight as to why they were designed that way. The
fundermentral goal which led to the evolution of Internet is develop a
technique
to interconnect different communication networks. The papers details seven
top level
goals, in the decreasing the order of importance. Firstly,
goal communication must continue despite loss of routers led to
stateless packet switches with more trust on host machines rather on the
underlying network system. Secondly, supporting multiple types of comm
service this
led to the idea of datagram division of intial single protocol into
two protocols: IP(connectionless or datagram) and TCP/UDP. Thirdly, idea
of supporting
accomodate different internetworks led to success of the internet itself.
Other 4 goals (permit distriubted mgmt of resources, being cost effective,
new host addition should be relatively easy for scalabilty,
accountability) though
not as important as the initial three contributed to the success of the
internet.

The author delves into the architecture and implementation of the
Internet. He
suggests several design aids such as protocol verifiers and simulators to
explore
new architectures. He strengthens the importance of understading datagram
as a
buidling block of Internet. The author critques that internet architects
in going
with the bye level flow control instead of going with both byte and packet
level
flow control. EOL field issues is also detailed.

The paper finally concludes because of the initial design goals - some
items such
as accountablity and resource management were not given lot of importance
due to
that it was initially designed for US miletary purposes etc. He also
quotes the
importance of flow(sequence of packets) being treated as a building block
as
that would have allowed designers to achevieve all goals effectively.

He initially
starts saying that ideas such as datagram service was never thought to be
so
important but later those characteristics themseleves turned out to
defining
characterestic of TCP/IP protocols. A good example of this is idea of
datagram
or connectionless service.
Received on Thu Sep 14 2006 - 11:12:28 EDT

This archive was generated by hypermail 2.2.0 : Thu Sep 14 2006 - 11:16:14 EDT