A Protocol for Packet Network Intercommunication
Cerf and Kahn

Review by Michael Cafarella
CSE561
October 6, 2004


Main result:
Cerf and Kahn present their gateway and TCP structures.
It's clear how these can be used to communicate in a 
least-common-denominator fashion between very different 
indvidual networks.


Strengths of paper:
It's hard for the paper to escape from the shadow of TCP/IP.
But it does what you expect: discusses how different networks
handle packet transmission, resizing, and addressing.  It also
talks about how TCP at the endpoints: ensuring correct ordering,
reissuing dropped packets, and port-addressing.  It's got all
the basics that TCP is known for.


Limitations, other problems:
There's not much on performance, which I think was a much bigger
issue with TCP in the early days.  Other blank spots are consistent 
with the priorities from the Clark paper, e.g., debugging, 
instrumentation, accounting.


Possible improvements:
Some support for the above areas would be nice, but is maybe 
asking for too much.  It seems safe to say this was a successful
bit of research.


Modern relevance, future work:
Most of the computer industry.  Much of this class.