From: Andrew Putnam (aputnam@cs.washington.edu)
Date: Sun Oct 03 2004 - 16:23:56 PDT
The Design Philosophy of the DARPA Internet Protocols
David D. Clark
This paper discusses the motivating factors behind the design of the
Internet architecture, and how those factors guided the architectural
development. The two key motivational factors were the need for
effective utilization of existing networks, and the need for a robust
architecture.
The primary factor driving Internet development was the need to
effectively utilize the existing ARPANET and ARPA radio networks. A
more elegant solution may have been possible by starting from scratch,
but the military had invested heavily in the existing networks, and the
Internet may not have been relevant had it been developed from scratch.
By using existing networks, the designers had some constraints to work
within. What the author does not acknowledge is the testing and
application infrastructure that is already existed thanks to these
networks. Evaluating the effectiveness of the Internet architecture was
much easier since anyone who used ARPANET or the ARPA radio network
could use the Internet architecture and provide feedback on the
effectiveness of the new architecture. The UNIX operating system became
a huge success not because it was the best operating system, but
because the designers used it every day, and had to make improvements
in order to make their daily lives easier. The same was true for the
Internet. The daily use made improvements necessary. If it were not
used, then the motivation to constantly improve the architecture would
not have been there, and it is unlikely that the resulting architecture
would have been anywhere near as solid as it is.
The second important factor that drove the Internet architecture
development is the military context in which is was meant to operate.
This requires the network to be robust since any node on the network
could literally be blown up at any moment. This robustness is arguably
the most important quality of the Internet as we know it today.
One interesting consequence of the development of the Internet
architecture primarily in the context of military networks is the
relatively high level of trust that the architecture places in the end
nodes of the network. The end nodes can affect the network by acting
maliciously, but this was not a primary concern for the architects. If
it were, there would have been more accountability built into the
architecture. As it is, it is very difficult to track an individual
that is adversely affecting the network.
The Internet architecture seems to be incredibly sound, but there are
several places where the author mentions that the architecture could
have been improved. For example, the author says that they should have
provided support for both packet and byte acknowledgment. I wonder if
these improvements have been incorporated into more recent TCP and IP
versions.
The relevance of the Internet today, as well as TCP and IP are obvious.
The paper includes an interesting note about changing the basic
building block for future Internet architectures to include state
information within the network that is not critical to the connection,
but would assist the network in providing resource management and
accountability.
This archive was generated by hypermail 2.1.6 : Sun Oct 03 2004 - 16:24:02 PDT