Paper #: 1.39
Title: The Web as a graph: measurements, models, and methods

1. Problems
-This paper tries to give a little order to the chaos, by applying some kind of mathematical model to the Internet. Previous models were merely
randomly connected graphs, but the authors try to prove that random graph algorithms do not apply to the Internet, and come up with an appropriate
model with algorithms of their own, in order to advance the mathematical study of the web.

2. New Idea and Strengths
-The idea of throwing out lower degree nodes (pages), creating a large base of nodes by expanding forwards and backwards on the first 200 nodes
returned from a simple query search, then finding connected subgraphs within the expanded base.
-The idea of pairs of hubs and authorities was suprisingly accurate, according to their testing. I was impressed by the large percentage of
"communities" they discovered when running through their elimination and generation sortings.
-Models for growth were clearly exponential under this model. Their ideas for modeling web growth seemed very interesting, with nodes creating copies
of existing edges, as edges are also randomly deleted and formed.
-The idea of looking for paths of alternating inward / outward edges was interesting, and pleasing that it was a fairly constant factor of 2 compared
to shortest length paths.

3. Weaknesses and Extensions
-I would like to see testing done for comparisons of alternating inward / outward edges, varying the number of times the same direction can be
travelled. To explain further, most paths are probably not as clean as hub to authority to hub and so on, but could be an inward to inward to outward-
the lengths of paths as they relate to the number of times an edge is traveled in the same direction could be interesting.
-I was a little disappointed that the original base node collection was composed only of the inward and outward nodes of the original 200. A graph of
the size of the base as it compares to the number of in and outward iterations could help show further the interconnectivity of the graph. Going an
extra iteration should find substantially fewer new nodes if the connectivity is as strong as they suggest.
-I thought the decision to treat intra-site links as nonexistant was not a great one, unless they took it a step further to treat the connected site
as a single node. It seems that two pages on the same site that were linked to each other will be treated as two independent, unlinked nodes according
to their thinking, which doesn't really make sense.