one could claim that this wasn't a bug, but an optimization.
that argument has legitimacy
even if it's a bug, what is the result? an almost immeasurable
impact on cpu. and, more importantly, absolutely no effect on
route selection
but when the behavior was discovered and cisco was told, they
changed it
so now it's up to the providers to deploy the code. if i were
a provider, then i might deploy it if i had more than just this
reason, but i very seriously doubt i would bounce a router for
this feature alone .. it's just not worth it
while i think it's a good thing in general to get rid of
excessive withdrawls, i think this thread is a good example of
us chasing numbers in statistics, purely for the sake of the
numbers, without really thinking about their significance
/jws