Hi Geoff,

In the "Re: [rrg] draft-narten-radir-problem-statement-05.txt" thread
  http://www.ietf.org/mail-archive/web/rrg/current/msg06152.html

and in "BGP in 2009":
  http://www.potaroo.net/presentations/2010-02-01-bgp2009.pdf

you presented some interesting research about DFZ activity.  I think
we are all greatly indebted to you for your long-term, BGP data
gathering and for your detailed and continuing analysis.

(Unfortunately, your mail client sends each paragraph as a single
line, so the web-archives require a copy and paste to an editor in
order to read them.)

I agree with your slide 30 comment:

    BGP Scaling and Table Size
      As we get further into the IPv6 transition we may see:

        – accelerated IPv4 routing fragmentation as an outcome
          from the operation of a V4 address trading market that
          starts to slice up the V4 space into smaller routed
          units

        – parallel V6 deployment that picks up pace

However, I think increasing growth in the number of IPv4 DFZ prefixes
will increase, for the reasons you suggest, irrespective of however
rapidly IPv6 is adopted.

It is puzzling that there is continued growth in the number of IPv4
DFZ prefixes while the rate of updates remains static.  Perhaps an
increasing proportion of end-user networks and ISPs are keen to avoid
being mentioned in your BGP Hall of Shame, AKA "The BGP Instability
Report":

   http://bgpupdates.potaroo.net/instability/bgpupd.html

I think your work is an excellent analysis of the most visible part
of the routing scaling problem - analogous to the part of the iceberg
which is above water: the burden on the DFZ control plane due to PI
prefixes and due to unreasonably frequent updates to a small subset
of all advertised prefixes.

Keeping a lid on these problems is important.  However, the
harder-to-measure part of the routing scaling problem is the large
number of end-user networks which want or need Portability,
Multihoming and/or inbound TE but can't get them due to the costs and
other barriers involved in gaining and advertising PI space in the
IPv4 DFZ.

Brian Carpenter recently suggested we should expect something in the
order of 10^7 such networks:

  http://www.ietf.org/mail-archive/web/rrg/current/msg05801.html

     Firstly, I'd like to support the ten million estimate. Based on
     the projected world population of just under ten billion by
     2050, that amounts to one enterprise per thousand humans, and
     that seems to be roughly what you find in the available data for
     developed economies (ignoring dummy companies and sole
     practitioners).

I have made the same 10 million WAG on the same basis.

This is not counting mobility, which could easily go to 3 or 5
billion hand-held devices in the foreseeable future, and perhaps as
high as 10 billion in the next century.  Some people exclude mobility
from the routing scaling discussion, but mobility is prominently
mentioned in the RRG's Charter.  With the TTR Mobility architecture:

  http://www.firstpr.com.au/ip/ivip/TTR-Mobility.pdf

a Core-Edge Separation architecture such as Ivip or LISP can be
extended to provide a globally mobile IPv4 address or IPv6 /64 to
mobile nodes, no matter how they connect to the Net (including behind
NAT or on scalable "edge" addresses) - with no mapping update
required when the MN gains a new address or access network.

I just wrote a message "Recommendation suggestion from RW" which
summarizes all the things I believe are important about scalable
routing and the opportunity we have for making major architectural
upgrades to IPv4 and IPv6.  Its nearly 10k words - but I hope you
would find it interesting.

  - Robin