[ppml] Incentive to legacy address holders

[Stephen Sprunk]

Thus spake "Leo Vegoda" <leo.vegoda at icann.org>
> I've not done the research myself, but I remember Harsha
> Narayan's 2003 research indicated that 89% of routing table
> prefixes were down to what he called splitting and spawning,
> not RIR allocation practices.
>
> http://www.ripe.net/ripe/meetings/ripe-45/presentations/ripe45-
> eof-harsha/page38.htm

Interesting work.  However, there's one new tool we've given (or at least 
tried to give) ourselves in v6: uniform prefix lengths.  If everyone filters 
the LIR blocks at /32, until an LIR qualifies for more (which should be 
extremely rare, and not any time soon) than that their allocation can't be 
split or spawn more-specifics.  Hopefully, what we'll see is people 
announcing a covering route plus more-specifics (for TE) that are filtered 
at N hops away where they don't matter.  Ditto for end-user blocks being 
filtered at or near /48.

So, we should see close to one globally-visible prefix per AS.  If (when?) 
we get to a million routes in the v6 table, it should be because we have at 
least half a million ASes.

That isn't possible in v4 because of the conservation requirement that 
dictates ISPs getting a new block every 6 mos, and the splitting and 
spawning that aren't feasible to stop due to varying prefix length.  That is 
why many networks are advertising an average of 14.5 v4 routes each, and 
even the majority of origin-only networks (i.e. leaf sites) announce more 
than one v4 route.

S

Stephen Sprunk      "Those people who think they know everything
CCIE #3723         are a great annoyance to those of us who do."
K5SSS                                             --Isaac Asimov