[arin-ppml] Taking back UNUSED addresses.

Geoff Huston gih at apnic.net
Sun Sep 14 17:26:54 EDT 2008

On 13/09/2008, at 7:42 AM, William Herrin wrote:

> On Fri, Sep 12, 2008 at 4:20 PM, Leo Bicknell <bicknell at ufp.org>  
> wrote:
>> In a message written on Fri, Sep 12, 2008 at 04:06:45PM -0400,  
>> William Herrin wrote:
>>> In other words, 100 - 62.8 = 37.2% of allocated space is not  
>>> routed on
>>> the public Internet.
>> Two other people have e-mailed me the same thing.  I believe you are
>> correct.
>> However, this effor is in the direction that puts more weight  
>> behind my
>> argument.
> Do we have an independent confirmation of the numbers? As I recall,
> that particular computation is not the primary purpose of that report.

I'll be brief - this list is already very busy and I'm reluctant to  
add to the reading volume unnecessarily.

The figures Leo Quotes appear to be from a weekly report generated by  
a BGP analysis script set up by Philip Smith of Cisco.

Another view of the advertised / unadvertised ratio can be seen at  
Figure 13 of http://www.potaroo.net/tools/ipv4/ (http://www.potaroo.net/tools/ipv4/fig13.png 
), using an independently authored set of tools that I set up that  
looks at the BGP routing table and the RIR's published data regarding  
allocated addresses.

This graph expresses the size of the unadvertised address pool as a  
percentage of the size of the advertised address pool over time.

The current value is 0.4322 (unadvertised addresses are 43.22% of the  
advertised address) Which means that of the total allocated address  
pool, 30.17% of the addresses are not _currently_ advertised into my  
particular view of the global routing table.

Its slightly lower than the figure that Leo quotes in the thread, but  
then again I can readily see that there are different methods of  
counting what's actually "allocated" and there are certainly scope for  
considerable variation here (see http://www.cidr-report.org/bogons/rir-data.html 
  if you want to have a feel for the quality of the input data on  
"allocated" addresses) and of course every view of the routing table  
differs to some small extent.

The current ratio has been stable, more or less, since early 2007 (the  
unadvertised address pool has been growing at the same relative rate  
as the advertised address pool in size)


More information about the ARIN-PPML mailing list