review1

Name : Gennady Pekhimenko
Student# : 995166225

  Summary.
This paper presents some ideas on how and where place functions in
a communication subsystem. These ideas and principles have got the name of
end-to-end arguments and they can help you not only with the design
of the computer system (or network), but also with a criteria on how to
create layered communication protocols. The main idea of them is not
to use functions at the low levels of your system, because they can be
redundant or not efficient. This happens, because we usually can implement
function completely and correctly only if we have enough knowledge
from the application. And our low level may be constructed before the
application creation. Even if we put this functions to the low level
and give them the property of having this knowledge, we will probably
get the performance problem. Mostly because this low level interfaces
are very widely used in our system and we will do a lot of redundant
work.

These end-to-end arguments were described for a wide group of
problems:
-- careful file transfer
-- performance aspects
-- delivery guarantees
-- data protection
-- message duplicating
etc. And for all of these problems we can use some kind of end-to-end
argument.

The paper is written well and has both general ideas with a good
practical examples for their using. The key idea I'd like to mention
is that we can make efforts to improve the lower levels, but "the lower
levels need not provide "perfect" readability". It is also a good idea for
using this method for layered communication protocols that just
started to appear when work was written (1984) and were in need of theoretical
background on how to divide the system on these levels.
As to the articles disadvantage I can only note the lack of any
measurements in the performance aspects part.
Received on Wed Sep 13 2006 - 22:12:58 EDT