At 19:38 -0400 8/10/00, Fred Baker wrote:
>At 01:33 PM 8/10/00 -0400, Corzine, Gordie wrote:
>>Wouldn't it be better by far, to assign new addresses from 000...1, and map
>>to routing information however we may code it?
>
>well, that is essentially what has happened in the telephone network, at
>least the wireless portion of it. Your telephone number is no longer
><country code><area code><exchange><house> but is essentially a random
>number which is used to look up your E.214 IMSI in a database, find the
>current topological location of the telephone, and then place a
>circuit-switched route to it. You move, the route follows you.
>
>What the Internet architecture tries to do is get rid of the
>circuit-switched route and the large-scale database. We do that by
>embedding topology information in the IP Address. This is fundamentally the
>difference between a connection oriented and connectionless network.
>
>No, it's not a stupid question. It's a paradigm question. You're asking the
>same thing Dave Mills asked in the mid-1970's: are we better off with
>circuit-switched routes or connectionless routes? Kleinrock's premise,
>underlying packet networking as we know it, is that the latter is a winner
>for survivability reasons. Those who form the bell-shaped side of the
>business tend to be of the other opinion. This constitutes something of a
>religious divide.
>
>As to the difference between the two, you can think of it in terms of
>telling someone how to get from Los Angeles California to Jacksonville
>Beach Florida. There are three ways:
>
> - I can give the guy a map and tell him to find his own way
> (that may be Active Networking)
> - I can tell him to get on Highway 10 and go east until he can't go any
> further
> - I can tell him to get on Highway 10, go east until he comes to a
>junction,
> and then ask someone for directions.
>
>The first is pretty complex when you think in terms of packet networks. The
>second is pretty simple, but gives the driver (packet) no options. The
>third is more survivable, but requires enough intelligence in the driver to
>ask the right question. The third is connectionless networking as
>implemented in the IP Internet. The topological address is what he needs to
>ask that question.
>
>Next issue is "so how do we get the right address?" - aha, something that
>you translate into a topological address for the purpose of routing is a
>name. The ITU even has a formal definition somewhere that puts it in so
>many words. We have names - DNS names.
