| Someone has questioned an assertation I have made and I want to make sure I'm not wrong before making an ass of myself. Picture a large network. Large enough that I have an entire public class B. There are lots of access switches, each of which connects to one of many distribution switches. Each distribution switch is further connected to two other distribution switches, as well as the main core switch and the back-up core switch. Each of the core switches has it's own connection to the edge router as well as a connection to the other core. With me so far? Nice, hierarchal partial mesh topology. So let's say my class B is 111.111.x.x The end-user wants to use RFC 1918 addressing on all the /30's in the core and distribution layers. Yes, we're actually worried about conserving address space with a full class B. No problem there. Additionally, in order to have some control over the router ID's, the end user has proposed that we make one core switch 223.255.255.254, the 'back up' core 223.255.255.253, and each of the distribution switches will also have an address decremented from the above. Still with me? It's much easier on a Visio... Anyway. There is no NAT in this network, therefore, when a packet comes from the access layer, through the distribution, through the core, to the edge router and the public internet, won't that packet pick up the router ID address? That is, when the edge router receives it from the core, it will think that it's coming from 223.255.255.254 rather than 111.111.23.65? I say, we have to either keep all addresses in the public or RFC 1918 space, or we have to run NAT. Otherwise, we'll never get to the public internet. Well...we'll never get back, actually, but you get the idea. Am I right or wrong? |
