When designing and implementing heuristics for discrete optimization problems,

there are many choices to be made. These include search paradigms, search

mechanisms, search parameters, test sets, etc. Often these choices are treated

as independent of each other, even though most researchers acknowledge some

interdependence between the diff erent search mechanisms and associated para-

meter settings. Our experience is that the search mechanisms and parameters

often interact, and at times in unforeseen ways. This is a diffi cult problem to

disentangle, and often even to discover.