WMI and NAT

The behavior you're seeing is inherently caused by the way DCOM is architected.    (And WMI is built on top of DCOM, so...).    When the client machine connects to the server machine via NAT in order to create the WMI DCOM object, the DCOM return packet contains a list of network addresses for the server machine.    The client machine will then try to ping the server machine via all of those addresses (IP addresses, DNS name, etc), and will use the first address that successfully responds.

In short, DCOM does not work through a NAT unless at least one of the NAT'd server machine's network addresses is visible from the client machine.    The best way to get this to work, is to somehow see that the DNS name of the server is resolvable from the client machine via the NAT machine.

Inability to translate network addresses within opaque network payloads is a well-known limitation of NAT and it shows up in other application protocols besides DCOM.

 From http://forums.microsoft.com/MSDN/ShowPost.aspx?PostID=102054&SiteID=1