used toillustrate an adaptive VPN

Source NAT. So far, in the cases we have used toillustrate an adaptive VPN, the selection of an appropriate tunnel has been based only on subnet policyrules, as it commonly is. However, an applicationbased adaptive VPN (as illustrated in Figure 6) alsoallows tunnel selection to be based on the destinationTCP port number. Unfortunately, routing tables specify routes by means of destination IP addresses and donot provide the flexibility to specify them by means ofa combination of destination IP addresses and TCP portnumbers. Let us consider again the configuration inFigure 10 and assume that now it is required thatpackets destined to the subnet 192.168.5.0/24 be sentthrough the enterprise tunnel if the destination TCPport number is 25 (i.e., e-mail) and through thenetwork tunnel if the destination TCP port number is80 (i.e., the Web). This means that the list of hostsbehind both the enterprise and network tunnelsshown in Figure 10 must be modified to include the subnet 192.168.5.0/24. The local presence IP addressesfor these two tunnels are 192.168.5.10 and192.168.1.10, respectively. But if a packet is to be sentto a specific IP address in the 192.168.5.0/24 subnet,there is no way to specify in the routing table that thegateway should be at IP address 192.168.5.10 if thepacket is to be sent to destination port 25 and at IPaddress 192.168.1.10 if the packet is to be sent toport 80.