Review for the Design Philosophy of the DARPA Internet Protocols

Name : Gennady Pekhimenko
Student# : 995166225

Summary

This paper gives as an idea on why we get the Internet like this and
what were the main reasons and goals when it was created. Because the
protocol suite for Internet - TCP/IP was developed by the DARPA
(mostly military project), few certain goals dominated the others.
Of course, the main goal is an effective multiplexing between existing
networks. There were two decisions: packet switching and circuit
switching. Most of the networks to be integrated were packet switching
ones and applications being supported used packet switching paradigm,
so, packet switching was accepted.
  As to the second level goals three ones dominated the others. They
were survivability in the case of failure, support for multiple types
of services and support for varieties of networks. Survivability
became so important, because of the military aspect of this project,
so we needed a possibility to work correctly even when networks and/or
gateways were failing. The second goal of these three was a possibility to
have a variety of types of services. The main requirements for them
were speed, latency and reliability. That's why appeared TCP -
traditional bi-directional service. Also there was made a decision
that more than one transport service would be needed, so it was the
reason of dividing one protocol into two ones (TCP and IP). The third
one - support for variety of networks was needed for the total
Internet success in incorporating military and commercial networks.

Because other goals had less priority, their impact on the design
was rather small. One of the paper's ideas is that change in this
priorities can greatly change your network.
The last parts of the article described such Internet's features like
datagrams and TCP and what decisions were made when they appeared.

I think that article is clear and well written. It shows as the "why
and how" of the Internet basis (TCP/IP, UDP etc) and gives an
opportunity to understand the designers choices. But it, for sure,
doesn't describe few modern complicated problems for computer
networks.
Received on Thu Sep 14 2006 - 02:12:55 EDT