I think the real difference is this. If my routes are only seen at one
place then it makes sense for me to announce them whether or not I can
actually get to them. If, however, it is possible to get to parts of my
network from one exchange, but not another then the routes should only be
announced where they can actually be gotten to from (obviously). I.e. if I
am a small to mid-sized provider who has multiple t-1's to the net to
different providers, and then those t's come into the same physical network
(either the same router or one's that are topologically adjacent to one
another) chances are good that if one of those routers can't get to part of
my network the other probably can't either. In this instance it makes sense
for me to null0 the routes and do my part to not affect instability (only
routes which originate in my AS though, if I have any multihomed downstreams
then they would not want me announcing their routes when I can't get to
them). If, however, I have a network which has multiple egresses to the
internet at large it is very possible for one part of my network not to be
able to reach certain sites while other parts can. I would then want to
"announce my instability" so that I am only announcing networks where they
can actually be reached from.
Justin "Long babbling posts R us" Newton
Justin Newton
Internet Architect
Erol's Internet Services