The Design Philosophy of the Darpa Internet Protocols
By David Clark

Review by Michael Cafarella
CSE561
October 4, 2004


Main result:
The paper doesn't describe an experiment as such, but does
discuss motivations behind the design of TCP/IP.  This is
pretty useful, as it helps explains why certain features
are present or missing.


Strengths of paper:
The strongest part of the paper is where the author discusses
the prioritized goals for TCP/IP.  These are brought back later
in the paper, to discuss why one choice was made over another.
The priorities here make a lot of sense, especially the point
that "accountability" was a low priority.  (There are still not
many good facilities for this, though perhaps that's because
network and computing resources are now so inexpensive.)


Limitations, problems:
It's a shame there isn't more performance data and quantitative
information.  The author claims that back-of-the-envelope calculations
were usually sufficient; that might be true for correctness, but
can't be true for performance.  It would be better to know the
costs of the features the designers added or removed.


Possible improvements:
I would have liked a rundown of every TCP/IP feature, along
with some history and motivation behind it (though this would
have blown up the paper size a lot).  In fact, this would be
a nice way to write the text.


Modern relevance, future work:
This paper is helpful for anyone who wants to try and improve
on TCP/IP.  It's also simply helpful for engineers in general, 
to see how a group of people went about one of the most successful
projects in decades.