[arin-ppml] Routing table growth, was Re: IPv4 is depleted today

[John Curran]

On Sep 3, 2008, at 10:42 AM, Iljitsch van Beijnum wrote:
> The 90% of the address allocations for small blocks (10% of the space)
> will simply continue in the same way they have for years, someone who
> needs a /24 will get a /24 and not four /26s. Or maybe they'll get
> _one_ /26. This won't impact the routing table growth much, if at  
> all.

Okay, I'm holding a very different set of assumptions...
where am I mistaken in the list below?

1. The majority of address space allocated today is going to a
    small number of very large ISPs.

2. These ISP's love to connect new customers to their networks, and
    have long-term financial forecasts based on this assumption.

3. Connecting a new organization to the Internet requires a small
    number of unique addresses (due to NAT use by many organizations)
    which can be out of provider space today (and hence routing neutral
    since they all come from the most recent very large RIR allocation
    to that ISP).

4. At depletion, ISP's will free up internal addresses and use them to
    the extent possible, and might encourage customers with PA space to
    give up unneeded address space in their assignment, but that's only
    going to allow continued growth for a fairly short time.

5. ISP's will then: A) Look at connecting new customers via IPv6 and
    the transition technology du jour, and/or B) Attempt to obtain
    more IPv4 address blocks.

6. If someone wants to 'make money' via address block transfers,
    they'll want to make the most money possible via such transfers.

7. The value of a /23 block is always going to be less than 2 /24's
    (since the 2 contiguous /24's are in demand to both anyone who
     wants the /23 *and* additional folks who desire just a /24).

8. The major ISP's realize that routing lots of "relatively small"
    address blocks obtained any way possible is really bad due to
    the routing table impact but other than (5A) above, it's the
    only option available for connecting new customers.

9. It won't take long for large ISP's to also realize that they
    are capable of affording 1M route DFZ routers, and such routers
    will let make use of even smaller obtained PI blocks and both
    handle the routes internally and share some of those routes
    to those other "tier 1"'s who happen to be doing the same...

10.  This model works "successfully" (although without any real
      of hierarchical aggregation due to extensive PI reuse) and
      provides us a small number of additional year of growth
      before collapsing due to departure from RFC 2008/BCP 7.

/John