Someone save me from my confusion.
I'm familiar with the classic use case of NAT/PAT, where internal hosts use many-to-one or one-to-one nat when going outbound. What I'm not as familiar with is what can I use if I have a 10,000 port range that external hosts request inbound to a host and I want my firewall to forward to that 10,000 port range to a single port.
inbound port range: 10,000-20,000
destination ip: 10.10.10.15 "internal host, public IP accesible IP"
translated port: 440
translaed ip: 10.10.10.15 "same as original"
source ip: any "internet"
Basically, port forwarding for a range of ports to a single port.
Where I think I'm getting caught at is the basics of tcp/ip where a client chooses a random port for its source port and sets the destination as the destination port, in my case 10,100. This hits my firewall with the destination port 10,100 and a source port randomly chosen by client, lets say 4570. The firewall would then say, I want to translate this 10,100 to the destination port of 440 and point it at my internal host 10.10.10.15. When the packet arrives at 10.10.10.15 it arrives with a source port of the now translated port 440. The firewall would not know how to get back to the original client that requested with source port 4570, thus my original request is not possible? This is where I'm at in my mind on why this doesn't work, but then again, I'm thinking that a PAT table could keep track of these connection mappings and broker the connection.
