The same friend who has asked me to give a presentation on why systems people should be interested in Mercury has also thrown down the gauntlet: the bet is that a Mercury implementation of the SPECINT 129.compress algorithm (LZW using double hashing) will not run more than six times slower that the C implementation; the wager is for about ten pints of beer or a couple of good nights down the pub. A stake worth playing for. [Just for the record, my friend would like to lose this bet, but is highly skeptical about the efficiency of declarative languages...]