At 10:57 -0700 2003.09.11, Bill Becker wrote: >It is not rediculous to expect a pseudo-random number generator to >behave differently! I believe that is, indeed what rand() is supposed >to be. You go in with a consistent seed across all platforms, it >should return consistent results across same. > >As I said before, the original poster (Louis?) should contact the >people who ported the windoze version of Perl.
rand() is not documented to return any particular results, as best I can tell. As such, I think expecting/demanding any particular results is unreasonable. Perl has long had different types of rand() algorithms on different platforms, often using one supplied by the system, and if one is not supplied, then some extra one. See all the config options in config.h/config.sh for RAND, including different functions, different number of bits, etc. For example, on Mac OS X, the Drand01() macro is set to call the system's drand48() (see man drand48). On Windows, however, it is set (rand()/(double)((unsigned)1<<RANDBITS)), where randbits is 15, as apparently no drand48() function is supplied by the system. I can't say why Windows is configured like that (see Config.pm for these values, I am reading from perl-5.8.0/win32/config*), but it is apparently intentional, and the porters of perl aer obviously well aware that different platforms use different implementations for rand(), which would necessarily result in different values returned by rand(). -- Chris Nandor [EMAIL PROTECTED] http://pudge.net/ Open Source Development Network [EMAIL PROTECTED] http://osdn.com/
