[ppml] Policy Proposal 2005-1: Provider-independent IPv6

Alexander, Daniel Daniel_Alexander at Cable.Comcast.com
Mon May 1 17:46:00 EDT 2006


I wanted to include some thoughts about the recent discussions. While
they are not necessarily the best ideas, they are sometime reality. The
two don't always go together. 

All the talk about not deploying PI space until the routing issues are
fixed, may not be very realistic. Unfortunately, resources are rarely
invested in a problem for the greater good. Money, time and development
roadmaps are defined by immediate problems, and any solution is not
going to be incorporated into code until long into the future. It is
possible, the routing issues of v6 will not be fixed until there is a
real world scaling issue, potentially due to PI space. At that point we
will have to live with the past mistakes. Necessity is the mother of
invention. We had our chance over the last 10 years to get it right and
we missed.

Talk about revoking or reclaiming allocations to correct the past is
unproductive. Once an allocation is made, it's done. To talk about ARIN
reclaiming space or making it "temporary" is an unrealistic solution.

Maybe the PA'centric aspect of v6 is a legacy design that we need to get
over. The business world has changed considerably since v6 development
first began and we need to adjust to it. There are many businesses that
have IT departments serving a larger footprint than many ISP. The needs
of these organizations must be considered and we need to adjust to the
future.

Many years ago, life was simpler. ISP's dealt with issues about the
Internet and provided services to customers. In today's world, many of
those customers have businesses that are fundamentally dependent on the
Internet. This means Executives do not want to be beholden to a
particular ISP. They require they be in control of their own fate. This
is why multihoming has become such an issue.  This has nothing to do
with technology and is entirely political, but they are not always
exclusive. A v6 world needs to provide for both models, PI and PA. Maybe
we are spending too much time arguing over one or the other, and need to
get started by embracing both.

-Dan



More information about the ARIN-PPML mailing list