On Fri, Jul 27, 2007 at 02:58:29AM +0900, itojun@itojun.org wrote:
so you know that "local optimization != overall success".
then why are you spending so much time trying to make incomplete
local optimization?
One could ask similar questions of BGP. Despite that, people find
BGP a useful tool, and while not perfect, it seems to remain bearable
for many network operators.
Personally, I'd find a mechanism for distributing address selection
policy useful. I have a number of sites that have both 6to4 and
native/tunnelled connectivity where the latter has uRPF applied.
I'd like our hosts to use all native addressing, unless they are
talking to a 6to4 host. Unfortunately, not all hosts have the right
default policy. If I could distribute a policy, they I could arrange
for everyone to have the right policy.
It's true that I probably can't get produce a policy to do the right
thing all the time, but that makes address selection no worse than
the routing protocols in use today. I guess your point may be that
we shouldn't have address selection at all, but I fear it may be
too late for that discussion?