On 2008 Oct 30, at 9:12, roger peppe wrote:
i'd be interested to know if you know of any studies on this.
i know of at least one system that uses it as the basis for its crypto. superficially it's certainly an attractive method, with minimal external dependencies, and, i'd have thought, at least a useful addition to just using the system time.
It turns out to be trivial to game randomness generated from scheduler activations: just launch enough other CPU-using processes to max out the scheduler and the search space gets *very* small. -- brandon s. allbery [solaris,freebsd,perl,pugs,haskell] allbery@kf8nh.com system administrator [openafs,heimdal,too many hats] allbery@ece.cmu.edu electrical and computer engineering, carnegie mellon university KF8NH