From: Ioannis Giotis (giotis@cs.washington.edu)
Date: Tue Oct 05 2004 - 19:29:51 PDT
A Protocol For Packet Network Intercommunication
VINTON G. CERF AND ROBERT E. ICAHN,
In this paper, we are presented with the original design of TCP/IP. The
authors describe in much detail a protocol that manages to connect
different packet-switching networks, addressing most (all?) of the
issues that could arise. Their main goals that are apparent throughout
the design are simplicity and flexibility.
This protocol is one of the very few examples in CS that have stood in
time. I believe the authors' main achievement is that by trying to keep
things simple they managed to make the protocol efficient and at the
same time able to support different transport or physical layers in the
future. This is mainly demonstrated by the ability to support different
packet sizes and their intuitive addressing headers. It should be also
noted that they're doing a fine job in justifying their demultiplexing
scheme in the process layer stage.
However, the design has its limitations. Some of the "hardwired"
decisions made are limiting factors in today's internet. The
retransmission scheme is not flexible enough to accommodate modern
streaming applications, and it is obvious that this protocol is not an
optimal choice when bringing wireless networks into the picture.
Clearly, the authors did not have an idea as to where things would go as
they specifically state-now obsolete- assumptions about the
retransmission scheme and its use, the possible size and number of the
networks or the protocol layer needs. No one can really blame the
authors for such decisions in that point in time.
On the other hand, it is surprising that a more discrete distinction
between transport and IP layers isn't made. The authors seem to
understand the concept of layering but they chose to incorporate a lot
of the protocol needs under a design that does not make clear (at least
in the paper) where one layer ends and where the next one begins.
Overall, a study of papers like these is extremely useful as we get to
understand better the key concepts that should be kept in focus, if one
wants their work to still be around after 30 years.
This archive was generated by hypermail 2.1.6 : Tue Oct 05 2004 - 19:29:52 PDT