Errata for 'Removal Policies in Network Caches for World-Wide Web Documents'

Abstract

Due to an error in the simulation used to conduct removal policy experiments, one of the primary conclusions from our SIGCOMM'96 paper is in error. We present here the corrections, and some new results. The conclusion 'sorting by SIZE is the best policy to maximize hit rate,' is correct. However, the other primary conclusion 'sorting by SIZE is the best policy to maximize weighted hit rate,' is incorrect. In fact, sorting by SIZE was found to be the worst policy. A brief description of changes in each of the four experiments conducted follows. The text of the revised paper can be found at http:-//www.cs.vt.edu/≈chitra/docs/96sigcomm/.