Regarding strange announcements of prefixes in 173.0.0.0/16 and 174.128.0.0/16
These prefixes are being used in academic routing research experiments. They are experimental space from ARIN, and we have paid the experimental fees and all that. Otherwise, ARIN is completely unrelated to this work.
The prefixes from this space are actually being announced by AS 3130 in the Westin colo in Seattle USA. But they have very strange AS-paths. We are using AS path poisoning (inserting others' ASNs in the path) [0] to keep those specific ASNs from seeing the test prefixes.
This particular experiment is to gather data for a journal/conference paper we hope to submit. It's SIGCOMM season, the paper dreadline is the end of January 2009, but the experiment could run into February [1].
Joe Provo pointed out that when Lorenzo was using this hack, some routers had 'problems' with long AS paths. This experiment is not using long paths. Lorenzo did so to get a lot of ASNs done in one whack. To get high effective throughput, this experiment slices two /16s into /24s and therefore can afford to put very few test ASNs in each path.
No real or production prefixes or data packets are being harmed in this experiment. To quote the most excellent Majdi S. Abbas:
randy lied but no packets died enough now
This experiment is being conducted by
Randy Bush <randy@psg.com>
and
Olaf Maennel <olaf@maennel.net>
2009.01.13
[0] - Lorenzo Colitti invented this hack and first published it
in early 2005. See his PhD
thesis and his
futile warning to NANOG, which produced no light but much heat,
sigh.
[1] - Due to route flap dampening, we can not change the
announcements as frequently as we might like. So it would be
prudent to vary a few test parameters, e.g. length of time an
announcement is up, to be more confident in our results.