From: Tom Christiansen (tomchr@ee.washington.edu)
Date: Wed Nov 03 2004 - 00:34:44 PST
This paper deals with "path inflation". The term refers to routing paths
being significantly longer than they should be. While the optimal routing
path is not necessarily the shortest path geographically, Internet routing
paths have been observed to occasionally be ridiculously long. This paper
investigates the issue and concludes that although ISP boundary policies
play some role, the most significant improvement in routing path length
(measured by delay) could be reached by optimizing the routing paths within
the individual ISP's networks.
There must be some rule dictating that SIGCOMM articles must be 12 pages
long and printed in a font size barely larger than the optical resolution
of the human eye.... ;-)
As with many of the other articles, I find this article to be longer than
what seems necessary to bring the point across. Some of the experiment
setup description could maybe have been put in a table rather than
described in the text. The methodology describes the experiment setup well,
but doesn't present many arguments for why this setup was chosen. Similarly
in the conclusion it is mentioned that the ISP's should be encouraged to
change their routing policies but I don't see any suggestions as to how
improve the existing routing strategies.
I would have preferred if the article had been written in third person
rather than the first person used. I could just be old fashioned, but third
person is more professional IMO.
I'm not sure I got very much out of this article. I look forward to the
class discussion.
This archive was generated by hypermail 2.1.6 : Wed Nov 03 2004 - 00:34:57 PST