Re: Measurement proposal
Neal Cardwell (cardwell@cs.washington.edu)
Mon, 18 May 1998 01:08:31 -0700 (PDT)
Seems like another class of sites we'll want to traceroute to is the set
of hosts where one could conceivably stick a wart, and then tunnel through
that wart - potential "virtual routers", so to speak. These hosts are the
ones that it wold be reasonable to have on the intermediate nodes of the
polygon: universities, research labs, ISPs, NSPs, and other entities that
could conceivably be convinced to carry transit traffic. It wouldn't be
reasonable to have AOL users in this set for example, because we'd never
tunnel through someone on a modem (or at least i'd be pretty shocked if
someone got better connectivity by doing so!). But we might have nodes in
this set that aren't web servers or clients - routers and misc other hosts
in the middle of an ISP/NSP cloud. How we get at this class of nodes, i'm
not sure. Perhaps we could consider every node along an internet path
(each hop in a traceroute result) to be one of these nodes. Then we get
traceroutes to them for free. Still, we may want to make extra efforts to
seek them out, because presumably there are a lot of critical routers in
the middle of the country (Chicago, St Louis, Cheyenne,...) that don't
have many popular servers or large pools of clients nearby, but play a
huge role in routing packets between the coasts.
neal