First, keep in mind that using Network Address Translation (NAT) you can re-use addresses, as demonstrated with the countless 10.x.x.x and 192.168.x.x network in-use today.
We're not going to talk about that.
Perhaps some discussion on the mask may help, don't focus so much on the address for now.
I'm sure you've read by now that the mask defines which part of the address is "network" and which part defines "hosts."
When describing the "networks," according to the mask as represented by the binary version of the value, each bit of the (binary representation) of the IP address that has a corresponding mask bit set to a "1" is considered the network portion of the address ... nothing new here, right?
When a device gets information coming down the (OSI model) stack, it looks at the destination IP address, and compares it to the mask defined for the egress interface.
If the destination address matches the network portion of the device's IP address, the device decides that the traffic is destined for a "local" host (same segment / broadcast domain as the transmitting device).
If the destination address doesn't match the device's network address (according to the mask), the device decides that the destination address is not local, so the traffic should be packaged in a layer two frame and sent to the Default Gateway (also called "Gateway of Last Resort" for Cisco routers / L3 devices).
The point, at this point, is to recognize that the mask is critical for the device to make the "local" / "not local" decision as to where it should send the frame (direct to the destination host or the Default Gateway, which should will forward the packet to the next network segment).
Using the same "host" IP address with two different masks means a couple things can happen:
1) Both addresses can point to a host
2) Both addresses could point to a "network address"
3) One address will be a host address and the other will be a network address.
In item 2 and item 3, the packet will not be forwarded properly (if at all - "Invalid IP address"), because a device cannot send a unicast (one source, one destination) to a network address.
In item 1, you will successfully get bidirectional communication ... but only for the range of addresses contained in the address with the longest network mask. Once the host number increments above the mask into the network portion, that device/host will think the address belongs to a different network and send the traffic to the Default Gateway (who will throw it away or forward it to the next hop ... depending on the mask for that interface of the Default Gateway).
Anything you may have read referring to route aggregation or "supernetting" only refers to network addresses, not host addresses.
It is possible (in fact, desirable)to use a single NETWORK address to represent a group of network addresses handled by a given router/Layer 3 device ... but there is no way, using unicast, to accurately define a complete block of hosts using more than one mask for a given segment; the device with the longest mask will define the size of the host address block that will work successfully for the segment.
NAT works because some gateway device has a map of inbound / outbound addresses to corelate ... there is no NAT on a single segment.
As for reading, here's a link to Cisco's INternetworking Guide online. It's a very good reference for dscribing the various technologies and protocols.
http://www.cisco.com/univercd/cc/td/doc/cisintwk/ito_doc/index.htm
Good Luck
Scott
