If, once you've found the machines, you want to run some sort of inventory system upon each of them, it sounds like you're going about this the wrong way.

 Rather than have one 'client' detect each Unix server - then demand it reports over it's software list/hardware information (or whatever) it seems to make more sense to get each server to send it's details to a "well known" reporting machine.

 This is what I've done upon my machines at work. Each server has SSH setup to allow automatic logins to a lowlevel account upon a spare Linux box.

 Once this is setup it's trivial to write a perl script to install upon each server to write out the machines users/load/package details/etc to a text file, and use 'scp' to store it upon the reporting server.

 The reporting server can then read the details from all hosts, and summerize to a daily email, or present the data interactively with CGI scripts, etc.

 If I've misunderstood your question please feel free to comment..

This is a good idea and is in fact what we do on the windows side for some statistics. However, in addition to just finding UNIX system info we want to be able to "discover" servers. In other words we want to be able to find UNIX servers that haven't been properly documented over the years, so we can be sure we have a complete server inventory. Thanks for the feedback.

Mitch