Magnus Hagander <magnus@hagander.net> writes:
> 2010/1/18 KaiGai Kohei <kaigai@ak.jp.nec.com>:
>> �The random seed is initialized at BackendRun() with MyProcPid and
>> �the time of backend process launched.
>> �Then, PostgresMain() -> InitPostgres() -> PerformAuthentication()
>> �will be called, and this random() shall be the first call just after
>> �initialization of the srandom().
> Maybe switching to erand48() would make this better, and good enough?
Wouldn't help in the least. The problem is not the RNG itself but lack
of an adequately unpredictable random seed, and anything you do here
is unlikely to be more random than what we already arranged for.
regards, tom lane