[arin-ppml] Draft Policy ARIN-2013-6: Allocation of IPv4 and IPv6 Address Space to Out-of-region Requestors - Revised

[Owen DeLong]

>> As I said above, the numbers do not tend to move as quickly as you claim.
>> Names tend to be quite dynamic. Numbers tend to be fairly stable. If they
>> were not, BGP would have a much higher (and unsustainable) level of churn.
>> 
> 
> Most of my addresses (in my tiny little Class C) have moved less than 20
> feet in the last 20 years, all are still in the same building :-)  This
> is mostly of academic interest to me as I try to envision the future of
> the Internet.  But also my company is trying to get its foot in the door
> in IP-based telephony, so Internet addressing and routing policies are of
> enormous interest to my customers (big telcos), and I need to understand
> their issues.

As a general rule, if you move around within EU, or within North America,
or even within much of Asia, you will often keep the same IP address.

In some cases, you will get a new IP address when you cross a national
boundary.

Other than within Turkey or Russia, I am not aware of anyplace where one
can easily cross a continental boundary with a cellular device and not
receive a new IP address.

I think that traversals from one RIR region to another without renumbering
are, therefore, quite rare (Turkey and Russia are entirely in the RIPE region,
even the parts that are in Asia). 

Even if devices move without renumbering, I think RIR qualification for
mobile services would be based on the "service delivery address" which
is the registered service address for the device, not it's current roaming
location. Such is the basis for sales and utility taxes, for example,
where applicable in the US.

> Is the general consensus that a mobile device would more likely re-number
> itself as it moves around, rather than transporting its address with it?
> Even in the glorious (mythical?) future of identity/location separation?

Depends on the movement, but, again, we are talking about pretty gross
scales of movement in most cases when you talk about changing RIRs.
For example, one cannot get to any region other than LACNIC from the
ARIN region without crossing an ocean.

I don't believe it is possible to get to any other RIR region from AfriNIC
without crossing an ocean.

APNIC and RIPE also have a common land border.

> If so, then address location (or subnet location, since that is probably
> what would really be measured) would be much less dynamic and so less
> problematic than I envisioned. 

Exactly my point.

Owen