[ppml] /48 vs /32 micro allocations

[Paul Vixie]

> > block size is just not a factor.
> 
> ... but consider this:
> 
> 	so for that /48 that your are injecting into the routing system
> do you mind if i "spam like hell for 10 minutes" and forge the source
> addresses as a random spread from within your /48?

you won't be forging source ipv6 addresses in a random spread unless you're
going to stand there with the route pointing at yourself so that you can
speak tcp.  (saints be praised, TCP-ISS is now a random-enough number!)

> you get to eat all the bounces and complaints - right?  hope your
> bandwidth is big enough to swallow the fallout. :) in this context,
> size does matter...

either you're there to be your half of the tcp state machine, or not, and
in that sense it's likely to be safer to carve an unused /48 out of someone
else's advertised /32 and "look like a traffic engineering route" then it
would be to grab a /32 or /48 out of the tundra.  i predict that the world's
appetite for historical bgp announce/withdraws is going to go through the
roof now that this method of spamming has reached a kind of "tipping point"
and there's no end-to-end bgp authentication on the horizon.