No, think of this as a resolution step that happens in a matter analogous to DNS resolution, but for IP<->IP address translation. At the beginning of a session, a translation request is made to resolve to the logical address (and all IP addresses are considered logical at first, just like all telephone addresses are considered logical until they are resolved). The translation is made, and the physical IP address is cached and used for the session. Obviously, end stations do not request this translation today so it would first require a protocol definition. Then it require changes in edge network infrastructure and/or client software, depending upon where the translation makes the most sense. This change, if ever attempted, would be highly painful, just like LNP was. But if you are forced to go through the pain, might was well try to solve the resulting routing problem too. Prabhu -----Original Message-----

From: Joe Abley [mailto:jabley@automagic.org] Sent: Thursday, March 15, 2001 10:37 AM To: Kavi, Prabhu Cc: 'Hank Nussbacher'; Stephen Stuart; nanog@merit.edu Subject: Re: Statements against new.net?

On Thu, Mar 15, 2001 at 09:59:53AM -0500, Kavi, Prabhu wrote:

...

Look at how local number portability (LNP) works. Before the phone call is connected, a translation is made between the logical number and the actual number. The actual number is based upon geography, and consists of country-code, area-code, local exchange, and then physical port number. As a result, the routing tables in telephone networks are small. For example, if you are in the US and need to call the UK, the network only needs one entry for all telephone networks in the UK (plus a few more for redundancy).

This translation/lookup function is only necessary once per call in a circuit-switched network.

In a packet-switched network, it's required once per packet.

For this reason, number portability on the internet and in the PSTN are quite different problems.

Joe