From: Ioannis Giotis (giotis@cs.washington.edu)
Date: Mon Nov 22 2004 - 08:30:49 PST
Routing in wireless networks cannot be handled with conventional routing
algorithms as most of them assume a rather static topology. The authors
present a routing algorithm for this kind of networks.
Although presented out of order, the algorithm makes certain assumptions
about the network graph and information provided, then tries to route
packets using a greedy approach, only to switch to a perimeter walk when
stuck.
Greedy algorithms are known to perform well in models such as the one chosen
by the authors. They try to address all possible issues and present
additional simulation results to backup their arguments. Overall, I found
the paper well written in this part and it gave me the impression of a
"strong" algorithm.
However, when the number of assumptions is increasing, one should be
skeptical. First of all, they assume geographical information from a source
they don't know how to construct reliably, secondly they make
graph-theoretic assumptions about the network (such as planarity) which
don't necessarily hold. Finally, as we've seen in a lot of papers this
quarter, that try to address issues that only emerged in practice,
simulation results are not enough to guarantee good performance.
Although, the paper presents some good ideas and the algorithm has desired
properties such as scalability and low state information, it is one of these
papers that could serve only as a starting point to a series of papers, that
could possibly lead to actual implementation. It is evident that more
theoretical work and experimental results need to be done first, before
anyone is convinced that this algorithm deserves any implementation effort.
This archive was generated by hypermail 2.1.6 : Mon Nov 22 2004 - 08:30:46 PST