I'm currently switching a small network to a sane configuration.

The current IP allocation is a mess: DHCP clients are being allocated
addresses from a set of public IP addresses which we shouldn't be
using, while servers have static IP addresses from the five we're
supposed to be using. All access to and from the internet is via a
consumer-grade NAT box, the "outside" IP address of which is on the
same subnet as the "inside" addresses.

The new configuration will be using addresses in the 192.168.17.*
range, with servers continuing to have static IP addresses, and only
being available from the outside world via NAT.

There are three servers on this network. One of them, the file/web
server, is running MacOS X Server 10.2, so it's stuck with its current
address for all of eternity.

As part of the changeover, I've got a Cisco 877 DSL router to replace a
consumer NAT box and DSL modem. I could use the NAT box to let the Mac
server keep its current IP address while looking like it's got an
address in the new range, but during a trial run, Windows file sharing
clients couldn't connect to it. Is there a better way of doing this,
or do I need to split the network into a "Mac" network connected to the
Cisco box and a "Windows" network connected to the old NAT box?

Also, under the current system, computers on the network are unable to
access the company web server by name -- the name resolves to the
public IP address, and something goes wrong with the connection --
maybe the NAT box isn't translating to the internal IP address? Will
this continue to be a problem with the Cisco box?

Thanks,
Mark