King Paper Review

The paper reports on KING, as a tool for accurately and quickly estimating
the latency between arbitrary end hosts. Like STING, this tool is not
introducing additional infrastructure into the web and thus can be scaled.

The strength of the paer is heavily due to the authors' extensive
evaluation of their simple idea to show its accuracy againts its
simplicity. One thing I specially liked was the study of the
application-level latency which has not been done for STING (in Stefan
Savage's Paper). At the same time, I guess it is expected that authors do
a comparison between their method and STING, since their nature is the
same (in terms of exploiting current protocols -- TCP and DNS).

All in all, I have nothing to say about their technical claims, but rather
on their philosiphy in which KING is said to be scalable since it does not
add anything to the current infrustructure. I doubt whether this calim is
correct or not, since the main idea is to add a NameServer (C) to fool
other nameservers and if we want to deploy KING worldwide, we have to have
several such servers, otherwise we'll have singular point of failure.
These nameservers could be flawed by attacks and etc.
Received on Thu Nov 03 2005 - 11:08:47 EST