Summary

This paper brings out an important problem: achieving global stability
among local people with competing interests. Traditionally, Internet
designers, users or operators shared a common purpose of transferring
packets over the network. Now, this no longer holds. The important and
powerful players of the Internet are at odds with each other. A simple
example is ISP must interconnect but ISPs are fierce competitors. Internet
has been designed by designers but its behavior is no longer predictable.

The paper brings out fundamental differences between mechanisms of
engineering and society, and the players that populate our field. It
outlines the design principles to accommodate within the Internet
mechanisms of society as well as those of engineering.

The basic idea is to design systems that accommodate variation for
outcome. The paper claims to design systems so that design is modularized
along tussle boundaries and designers should design for choice to permit
different players to express their preferences. It also explains different
flavors of tussles and why tussles often occur across interfaces.

The paper discusses the aspects of tussle through examples and illustrate
the how the design principles can be applied in practice. For example,
providing fiber based residential broadband access is motivated by the
principle of design of choice and as well by the recognition of tussle
boundaries. Though it is good to define that protocols that have clear
tussle boundaries defined, sometime it may be difficult to define the
boundaries in advance. They evolve as the system grows. In fact this was
the case with Internet. So, I fell that it is important designers should
consider these principles, however, they may not have clear tussle
boundaries defined for all cases before that protocols are designed.

The author also calls for clear separation of policy and mechanism. True
value neutral design is difficult in practice. However, one of most
important design challenge is to discover the part of mechanism that can
be divorced from policy.

Some of the design concepts stated in the paper may conflict with the end
to end argument. For example, end to end arguments call for transparency
whereas loss of trust calls for less transparency.

I was impressed by the profound knowledge of authors. I found that
examples such competing ISP, namespace management in DNS, and QoS
management interesting in relation to the design principles stated in the
paper.
Received on Tue Nov 28 2006 - 08:01:23 EST