From: Shobhit Raj Mathur (shobhit@cs.washington.edu)
Date: Sun Oct 03 2004 - 22:11:53 PDT
The Design Philosophy of the DARPA Internet Protocols, SIGCOMM 1988
-------------------------------------------------------------------
Main Result:
-----------
The paper enumerates the original objectives of the internet architecture. It describes the early motivation and
reasoning which led to the design of this architecture and which shaped the internet protocols.
Strengths of the paper:
-----------------------
- The paper successfully explains why the intrenet protocol was originally designed to be a connectionless service
based on datagrams. Though there were many papers at that time which described the protocols and their working, there
were not many which discussed the reasoning and motivation behind them. This paper is unique in that sense.
- It enumerates and explains the design goals of the original internet architecture. Moreover these goals are listed
according to their importance. This gives the reader a better understanding of the underlying motivations. A
different order of importance would result in a totally different architecture.
- The paper points out a few deficiencies of the original design. It suggests a few changes to it (e.g. adding state
information to datagrams) which would enable it to cater to a wider set of users and also address some of the goals
which were lower down in the priority list.
- The paper also identifies deficiencies in TCP for applications like tele-conferencing and debugging, and the
motivation for protocols like UDP.
- The paper can be read by anyone having basic knowledge of computer networks and systems.
Key Limitations, Unproven Assumptions and Methodological Problems:
------------------------------------------------------------------
- The paper concentrates on a few primary goals of the internet architecture. It fails to give alternate designs to
meet the goals of lesser importance (e.g. cost effectiveness, host attachment, accountability). This is a minor flaw
as the intent of the paper was to motivate the original design.
- It also does not give a convincing explanation as to why packet switching was chosen. It merely compares with circuit
switching and uses remote login as an application to justify the choice of packet switching. The reasoning could have
been more sound.
How could the work be improved ?:
---------------------------------
A few lines could have been added to describe how the design can be
modified to meet the lower priority goals. Packet switching should have
been motivated in a better manner.
Relevance today and future work:
--------------------------------
It is amazing to know that the original internet design which was based on packet switching, TCP and datagrams is still
used today. Even 16 years later the underlying design has not undergone any changes though the applications are vastly
different. The original motivation to interconnect a wide variety of networks is still relevant today, where ubiquitous
connectivity is a concern.
This archive was generated by hypermail 2.1.6 : Sun Oct 03 2004 - 22:11:53 PDT