[ppml] too many variables

Edward Lewis Ed.Lewis at neustar.biz
Mon Aug 13 10:51:54 EDT 2007

At 2:08 -0700 8/13/07, william(at)elan.net wrote:

>Isn't the right place for this question NANOG or IETF?

Ordinarily I don't like to discourage discussion no matter how 
tangential it is to the main theme of the list, but in this case I 

ARIN has the responsibility to develop and follow address assignment 
procedures that maximize the value of the networks it supports, in 
particular the global public Internet (but not limited to that 
network).  To do this, it is important to understand, like an avid 
fan of a sport, the nuances of the network and in particular it's 
system of routing.

Like sports fans though, ARIN is powerless (realistically) to "fix" 
the problems of the operational aspect of the network.  No more so 
than yelling "change the pitcher" gets the manager out of the dugout 
and to the mound will our discussion of Moore's law help the 
situation.  Especially as there are so many subtle points to balance.

Reading though all of the messages on this list, I don't believe 
there is any policy ARIN can use to solve the problem of routing, 
hence, not remove what is probably the largest bottleneck to IPv6 
adoption.  (I recall a discussion about the about to started 6Bone 
experiment in which someone said "it's a great idea, but as it rides 
on tunnels in v4 we won't get to test out any meaningful routing 
solution.")  But we can scream "change the pitcher (protocol)" when 
the pitch count gets high.

IPv6 has more destinations than IPv4.  IPv6 was supposed to be able 
to count on traffic aggregation to make it run smoother.  In the 
intervening years the network has evolved in ways that make route 
aggregation less likely to happen.  Demand for traffic engineering, 
multihoming, etc., drive PI space needs.  Failures of ISPs post-boom 
have made organizations unlikely to rely on one ISP's address space.

I don't think that the routing problem is a situation ARIN can really 
fix.  ARIN can only hope to maximize the benefit of allocating 
addresses one way or another to make the networks work.  We can't do 
much about solving the need to reach all the destinations in the 
network in the way the users of the Internet want to reach them (and 
be reached by them).

It's good to be up on the topic, but what is apparent is either a new 
engineering approach to find one's way from one point in a network to 
another or some operational practice that divides the problem of 
navigation into simpler ones that we can solve with proven technology.

Edward Lewis                                                +1-571-434-5468

Think glocally.  Act confused.

More information about the ARIN-PPML mailing list