[ppml] 2005-1 or its logical successor

For the purposes of discussion, one approach would be to define the PI space
in a way that could be aggregated the further it went from the source. A
sequential assignment scheme creates a swamp that is hard to manage over
time. An approach like:
would allow for aggregation at distance, while still allowing those that
want to have a route carried globally to enter into a direct business
relationship with each carrier for that routing slot. This would seem to
align the burden with the financial model. 

Even if you started with the assumption that there was no aggregation in the
geo based PI approach, as it became popular in each region a business case
for exchange based aggregation would emerge. I know that 'topology does not
match geography', but this approach would act to constrain topology in a way
that would force alignment of the finances with the exceptions.