[R] random number generator?

[Bruce D McCullough]

With respect to the method used to produce
normals, Professor Ripley noted:


>Were it not for concerns about reproducibility we would have switched to 
>Inversion a while back.

Presumably this refers to reproducibility across platforms.  

I searched the archive and could find nothing on this point.
Gentle's text on RNGs provided no answer, either.

Might Professor Ripley or someone else briefly explain why the
inversion method is not reproducible but Box-Muller is, or 
perhaps offer a citation?

Thanks,

Bruce