"... In PAGE 22: ... The technique in which low freight EOL to EOL arcs not in the HLP are eliminated is now the only exible pruning method we have. In the rst set of tests performed in Table7 , this low freight parameter is set to a reasonably low value. In the second set of tests, the low freight parameter is increased so that all EOL to EOL arcs not in the HLP are eliminated.... In PAGE 22: ... The di erent settings for freight density and extra handling time remain the same. Table7 displays the results obtained by running the UND{based algorithm on the Averitt data set. The larger network resulted in a considerable increase in time.... In PAGE 22: ... It is particularly interesting to note that the solutions obtained using only the HLP arcs in the network have the greatest improvement in objective function value, and the worst performance in terms of handling and service violations. We infer that if a better pruning technique were available to us { one in which we could prune approximately 90% of the arcs not in the HLP { we could obtain even better solutions than those in Table7 . As the results stand, the solutions are quite good.... ..."

"... In PAGE 21: ... The technique in which low freight EOL to EOL arcs not in the HLP are eliminated is now the only exible pruning method we have. In the rst set of tests performed in Table7 , this low freight parameter is set to a reasonably low value. In the second set of tests, the low freight parameter is increased so that all EOL to EOL arcs not in the HLP are eliminated.... In PAGE 21: ... The di erent settings for freight density and extra handling time remain the same. Table7 displays the results obtained by running the UND{based algorithm on the Averitt data set. The larger network resulted in a considerable increase in time.... In PAGE 21: ... It is particularly interesting to note that the solutions obtained using only the HLP arcs in the network have the greatest improvement in objective function value, and the worst performance in terms of handling and service violations. We infer that if a better pruning technique were available to us { one in which we could prune approximately 90% of the arcs not in the HLP { we could obtain even better solutions than those in Table7 . As the results stand, the solutions are quite good.... ..."