From: Tyler Robison (trobison@cs.washington.edu)
Date: Sun Oct 03 2004 - 23:40:24 PDT
The paper provides some insight into the TCP/IP protocol suite by
looking at the initial goals in mind for the protocols. These goals were
each assigned a priority, with survivability being listed first, and the
resulting architecture was meant to satisfy these goals in terms of their
priorities, while still fulfilling the overall idea of connecting the
various existing networks.
As mentioned in the paper, there is some uncertainty &
misunderstanding as to why the protocols are as they are, and the paper
does a good job of clearing this up by explaining the train of thought
behind the architecture; the choices made make more sense after reading
this. For example, the idea of splitting one protocol into TCP & IP makes
more sense given the need to support both reliable and unreliable
delivery. And, more importantly, it seems to be a fairly objective
analysis, listing both the positive and negative results of the choices
made; for example, it states that it would have been better to have the
flow control based on both packets and bytes instead of just bytes.
In terms of limitations, the analysis of the lower priority goals
was pretty light, and it would have been nice to see some more alternative
possibilities and why they were discarded. On the other hand, the low
priority goals were fairly unimportant, and the paper was meant to explain
the result, not to chronicle the entire design process, so its
understandable that these were given minimal coverage here. Nonetheless
it would have made their decisions more clear if we could see some more of
the alternatives.
Fulfilling the last goal listed, involving monitoring resource
usage, seems like it deserves much more emphasis, and the author does
mention some ideas for the next-generation architecture which would help
with this. A more expanded version of these ideas would have been
helpful, though its understandable why it was not included in this paper.
The ideas here are still relevant today, given the prominence of
the internet. The particular philosophies behind it are hardly necessary
for most applications using the internet, but understanding such issues
could be helpful in designing and implementing any large scale network.
The paper also shows some of the strengths and weaknesses of the
protocols, which could be similarly useful. And certainly this would be
useful in thinking about 'next-generation' protocols, as mentioned in the
paper; learning from the solutions and mistakes of the past.
This archive was generated by hypermail 2.1.6 : Sun Oct 03 2004 - 23:40:24 PDT