I'm running a configuration of this sort myself: we have an old (PII-266) HP Pavilion that we're using as the firewall, with two NICs installed. The Internet-facing NIC is configured using the IP address from our ISP, and is connected to our DSL modem. The inward-facing NIC is set up using the test IP subnet 192.168.2.*.
The Pavilion's running RedHat 7.3, with a custom kernel built for MASQ/NAT (IP Masquerading/Network Address Translation) and NetFilter support. All seems to be running well. We're using FireStarter on the firewall machine to set the firewall rules for it to use, as well as for monitoring what kind of requests are coming in (and getting dropped to the floor) from the outside (http, ftp, netbios queries, etc). It's been running like a top that way (even with OS/package updates) for close to a year now.
Even with a configuration of this type, you still want to ensure that the hosts on the inside are configured for minimum network services and maximum anti-virus support, as these are the principal entry vectors for exploits. In checking your configuration, have a friend on the outside, somewhere on the Internet, perform a scan of your firewall box by IP address, using a tool like Nessus, to determine if you have any exploitable holes open.
Overall, I highly recommend the configuration. Best of luck, and if you need other pointers, please feel free to post them on this thread, as I will be reviewing it pretty regularly.