On Tue, 25 Feb 2003, gabriele renzi wrote: > It seems to me that I noticed something about this thing in your web > site. > Could we hope you just did some work and could integrate it with the > actual MT implementation? > Yes, most of the needed code is in RandomR. Needs to be integrated with random.c in ruby src though. > Having a PRNG class would be cool, and marshaling ( seen has > seed-saving) would be good too. > It would even be cool to have a choice, cause (I seem to remeber) > that MT is not a *secure* algorithm. > > And.. this hypotethical module could support underlying OS's RNG > (at least on some ) .. > We could benefit of powerful alghoritms on modern OSes, such as Yarrow > , and even get the fast native implementation. > Yarrow is much too complex for the ruby src distribution imho (needs lots of code for crypting state etc). But sure, a natural consequence of objectifying mt is that we could make it pluggable (ie rand calls $DEFAULT_RNG.rand and srand(s) does $DEFAULT_RNG = $DEFAULT_RNG_CLASS.new(s) ). Regards, /Robert