That is what I always do.

You get a sandbox, to play in. Never will you ever get privileged access, unless specifically given. A combo of the tools that exist currently for UNIX in general gives the proper access. For one thing CVS pserver should NEVER be run as a "trusted" user... especially if it has public access. Not a single machine I have setup for production has been able to hammered by a user... except Oracle gobbling up all the Swap Space...

Now my personal machines, I really could care less if you hammer them. They aren't production... (BTW, Scott that rootkit and pr0n server works well...)

Testing machines are just that... testing to see whatif? In the whole time I have been doing this, I have only had 2 machines cracked into. Of the many I have setup. One was before anyone REALLY understood what non-Military firewalls *SHOULD* do... and one when I had just gotten done installing and updating the system... and had installed the base applications and forgot to turn off echo, daytime, etc... and rsh and rcp... oops. Classic exploits waiting to happen. Caught them *ON* the machine. With a root user named "aolsrv" which blended in with the "aolsrvr" untrusted user I was using. Of course, the network cord seemed to fall out of the machine at that point...

But, as the case may be, I call this way of thinking... "(l)usermode" therefore *I* can still get to the machine and fix it, with out it being *SO* hammered I can only hit the big red button. I believe others call it this too...