[ppml] Soliciting comments: IPv4 to IPv6 fast migration

[Owen DeLong]

On Jul 27, 2007, at 12:08 PM, Ted Mittelstaedt wrote:

>
>
>> -----Original Message-----
>> From: ppml-bounces at arin.net [mailto:ppml-bounces at arin.net]On  
>> Behalf Of
>> William Herrin
>> Sent: Friday, July 27, 2007 9:14 AM
>> To: Keith W. Hare
>> Cc: ppml at arin.net
>> Subject: Re: [ppml] Soliciting comments: IPv4 to IPv6 fast migration
>>
>>
>> On 7/27/07, Keith W. Hare <Keith at jcc.com> wrote:
>>> With some amount of push from customers and lawmakers, the telephone
>>> companies have moved from Provider Agregatable phone numbers to  
>>> Provider
>>> Independent phone numbers.
>>
>> That's a great point Keith. And here's the nasty part: because they
>> waited until the issue was forced, they had to make it fully PI,
>> individual number by individual number. They lost the option to use
>> some sort of sensible grouping strategy.
>>
>
> I think we have carried this analogy to the point of silliness.
>
> Area codes still create groups.  But more importantly, the phone

Really, they don't.  Area codes are portable across multiple providers.
There isn't even a separation of area codes between Cellular and
Local providers.

> number can be an abstraction because it is only used 1 time during
> the call - at the beginning for the phone switches to setup the
> call.  Once that is complete and the query into the lookup table

That's true, but, only mildly relevant.

> that matches the PI phone number to the internal routing number used
> by the phone company  is complete, the table isn't queried again.
>
Much like mapping of domain names to IP addresses.

> With IP traffic, to implement something similar to a PI IP address,
> you would have to have every non-edge router on the Internet make
> a query to a lookup table of some sort, and they would have to do it
> for every packet.  For a VoIP phone call that might have 10,000

This is not necessarily 100% true.  The router could cache the lookup
result for previously seen destinations.  Another option would be
to encode the destination AS on the packet near the edge and
route across the DFZ based on ASPath/Next Hop data without
the need to keep prefix data distributed.

> packets in the entire call that passes through the routers during
> call existence.  You can't do a query for each packet.  That is why
> IP is still going to require some sort of "sensible grouping"
> and why telephone numbers don't.
>
Actually, that's why IP needs a new routing paradigm more than
why phone numbers don't require grouping.  IP only requires
grouping if we continue to use the End System Identifier as
the Routing Locator.  There have been several proposals
for ID/LOC split.  In the long run, it is ID/LOC split that is needed
to make this feasible for both the phone companies (which
have it already) and for IP (which doesn't yet).

In the phone companies, the ID and LOC look a lot like each
other and are sometimes even the same number.  In IP, there
isn't really any benefit to doing so, but, we could do it that
way if we wanted to.

Owen