It is running in NAT mode.
It works fine for everything except connections initiated from the firewall itself. The comptuers on the LAN can connect to the internet, and are NAT' ed to the public addresses in our range 208.191.50.0/23.
For example:
1) Workstation on the LAN connects to www.google.com, the source is nat' ed from 10.x.x.x to 208.191.x.x, and routed correctly.
2) If the firewall itself attempts to connect to www.google.com (such as a ping from the console). The source address would be 172.20.1.2, which can easiely make it to the next hop 172.20.1.1, but that router does what is proper and drops the packet, because the source is not a publicly routable address.
This is a common scenario for router-to-router connections to be on private addresses so as to not waste public ip-addresses. When I was asked about this, I agreed. It seemed a good idea to me that connections from the internet could never be made to the firewall itself. I just neglected to think about the firewall making connections out from it' s own address. With my previous FortiGate unit I didn' t do any anti-spam, anti-virus, or web filtering, so the firewall never needed to initiate connections to the internet, only NAT connections as they passed through.
So...The oversight is on my part, but I was hoping there could be a solution at the firewall. The proper alternative is for me to pull out a /30 subnet from my public addresses and use that for the router-to-firewall connection, or to simply not use the web filtering and anti-spam features. I don' t need the anti-spam anyway and not sure if the web filtering will be of much benifit either, I just wanted to test it.