Brian Warner wrote: > So we can imagine somebody who's one of those rational economist types, > trying to make a judgement call. They just finished deciding to not buy > a lottery ticket (payoff*probability < price), and deciding to drive to > work (utility-gain > chance-of-accident), and to skip the skydiving trip > (metric-fun < chance-of-death*non-funness-of-death), and now they want > to decide whether they want their document-validating system to > fail-safe or not (i.e. prefer consistency or availability), and they've > got their utility function all set to go.
It's impossible to guarantee availability under any reasonable assumption. So you might as well get consistency, which is much easier, regardless of your utility function. -- David-Sarah Hopwood ⚥ http://davidsarah.livejournal.com _______________________________________________ tahoe-dev mailing list [email protected] http://allmydata.org/cgi-bin/mailman/listinfo/tahoe-dev
