[arin-ppml] Q1 - ARIN address transfer policy

[John Curran]

(context is IRTF RRG work for locator/eid based routing)
At 2:07 PM -0400 6/22/08, Joe Maimon wrote:
>...
>As I see it, hierarchical routing tables and transition to routing
>schemes other than hop-by-hop routing ...

Exactly:  "other than hop-by-hop routing"

If you can reach any point via a topologically derived locator (loc)
and can always translate a unique, non-topologically assigned
endpoint identifier (eid) into its current list of locators, then the
you're done.   Topologically-derived locators don't inherently
require additional routing information, except optionally for
purposes such as route diversity, traffic engineering, etc.

The problem is that the mapping of EID to its locators is a
wide open research problem, with several alternatives being
explored but no definite answers.  To quote Randy Bush:

   ' watch out, on the third step there
     is a sign "magic will happen here". '

Actually working on this problem is very important long
term, since the current model which combines semantics
of both EID's and locators doesn't scale long-term without
a strictly hierarchical PA allocation model which is not really
customer friendly.

/John