[ih] IP addresses are not phone numbers, was Some Questions over IPv4 Ownership

[John Curran]

On Oct 17, 2010, at 2:55 AM, Alan J Maitland wrote:

> John,
> 
> Good points and I really enjoyed your use of the term "peripatetic traveler".
> 
> Does anyone on the list know if there is a plan to more geographically allocate IP addresses in V6?  It seems like doing that might make it easier to control routing table sizes.  In other words, I'm going back to a section of one of John Curran's contributions to this discussion earlier talking about CIRs, which seem to flow down from RIRs (see http://www.scribd.com/doc/29487289/ARIN-Contribution-to-ITU-T-IPv6-Study-Group ).
> 
> What I gleaned from that link was that using the RIR to CIR structure looks a little like the hierarchical structure of DNS.  If portions of the V6 address space could be allocated in a way that was geographically aligned to the RIRs and CIRs and maybe even regional routes within a country by major carrier, would that not take some load off the primary routing resources?


Alan - 
 
  There have been discussions for geographic addressing in IPv6, but
  it doesn't improve routing (actually, the converse) unless the path
  of connectivity actually follows the geography.  In order words, it
  can have useful returns where the entire country has one mandatory
  ISP which everyone much utilize for their transit to the rest of the
  Internet.  In that circumstance, you can switch "local" ISPs and 
  have no impact on the global routing table since the global table
  has only a single aggregate for the entire country.

  The problem with this model is that it's completely divorced from
  typical reality, whereby numerous distinct international ISPs all 
  are vying for business in a country.  If you attempt to overlay a
  geographic addressing model, each ISP needs to announce slices of
  those geographic prefixes to the greater Internet.  This is indeed
  like what happens presently with local number portability, but one
  must remember that phone destinations are looked up via SS7 once at
  the beginning of the call, and the typical phone call lasts tens or 
  hundreds of seconds.  This yields an acceptable ratio of lookup 
  overhead in the circuit switched world. Attempting to do the same 
  in the packet network results in that lookup being performed for 
  every packet, as it arrives as every backbone router, globally.
  The imputed cost of a *single* non-hierarchical route in the global
  Internet has been estimated to be $15000 to $25000 worth of overall
  capital costs, so trying to avoid needless routes (or accurately 
  recover the associated costs) is a high priority.

/John