Paper Review, Di Niu

In this paper, the authors presented their end-to-end argument, as a
guideline in designing distributed computer systems. They argued that
different from data transmission systems, computer networks designers
should pay especial attention to higher level functions. They first
justified their thoughts of end-to-end caretaking by enumerating the
possible threats to a file transferring application. It turns out
through their arguments that to reinforce strict reliability on each
step of the file transferring application is too costly, and any
schemes that guarantee an error-free low-level data transmission
cannot guarantee end-to-end reliability. Based on the justification
of end-to-end argument, the authors then discussed the trade-off in
placing functions in different levels of the system Their discussion
is based on performance aspects. Their conclusions are that both
higher level and lower level functions are necessary and how much we
should investigate on each depends on specific applications. On one
hand, too much weight on low level functions could be costly, while
on the other hand, too little weight on low level functions could
degrade system performance greatly. At last the authors provide
historical context and examples of end-to-end argument such as
delivery guarantees, secure transmission, etc.

This paper provides a basic guideline for the layering design of
modern computer networks. However, the trade-offs among, transferring
speed, reliability and cost are so complicated issues that we need
carefully tune functions in different layers for specific
applications, such as wireless ad hoc networks. Such applications
specific difference are not seen in the paper.

Reviewer: Di Niu
Received on Wed Sep 13 2006 - 19:08:06 EDT