On Feb 20, 2004, at 2:57 PM, John Wroclawski wrote: > "construct a model of the system assuming it's all XCP, and then watch > for behaviors that are deviations from that model". (There are > probably others than just loss). Like delay growth?