[ppml] 2005-1:Business Need for PI Assignments

[Paul Vixie]

<hat="just another bozo on this bus" />

> Not only large institutions find it painful and expensive to renumber.
> Small ISP's take just as long, and relative to their size the expense is
> as great or greater than for a large one.  Remember staff resources are
> proportional to size of the organization.  Years ago it took my small
> ISP months to migrate from one provider's /21 to another provider.

in the ietf it is considered reasonable to focus on engineering and 
operational expense and to ignore business models when standardizing
something.  in the RIRs, business models actually matter.  the compelling
issue for small isp's is that if they live in PA space they're
effectively acting as a VAR for their upstream.  if you ask a small
isp in this position to renumber, they have to ask their downstreams
to also renumber.  these downstreams have to consider the alternative
of switching to the old upstream so as to avoid the current renumbering
event, or to switch to a larger (non-VAR) provider to avoid future
renumbering events.  this significantly weakens the competitive position
of small isp's and is bad for the industry and for the community.

in other words i agree with the text i quoted but my reasons are different.

> The current IPv4 minimum of a /19 or /20 is already tough for small
> ISP's to reach and eliminates nearly all end-user organizations (in the
> larger context of millions of end-user organizations).  I think the IPv6
> limit should target the same size of organization as the current IPv4
> limit does.  If I multi-home and can qualify for a PI block under IPv4,
> I should be able to get one in IPv6 too.

with the provisos that address space cannot be guaranteed or even known
by an RIR to be routeable, and that if/when the IPv6 routing table ever
grows large enough to be a problem then a new equilibrium point will
have to be found: i find this to be a sensible proposal.