Re: RFC: mt_* functions

[<jeroen@php.net>] > Hi, > > Currently, the rand_functions all have mt_ clones, which use a > in-PHP-implementation (Mersenne-Twister) rather than an external > implementation. > > This is IMHO a bit strange way of chosing between implementation. My > suggestion is to make it only one familiy of functions, the implementation > of which is determined by an ini-entry. (rand = system vs. rand = mt, or > something like that, default: mt). > > The functionality is the same (*), and IMO it's a per-system-decision wether > you want to use the mt-algorithms or the system ones. (usually mt, unless > your system has a real good / very special one) > > For backwards compatibility, mt_* can be aliased to their normal > equivalents. Please don't. Ini settings that change semantics are a bother, and people should be able to choose their random function. - Stig -- Stig Sæther Bakken <ssb@alltheweb.com> Fast Search & Transfer ASA, Trondheim, Norway