I now that my previous post is pretty wrong, but point is that this randomizing of bits is combined with random value.
EDIT: I know where i read about it! It was either in man pages or linux howtos, where they talked about using high word and low word in 'ancient' times, and that with "new" random() they don't have to anymore. It was few years ago.
EDIT2: from "man 3 srand":
'However, on older rand() implementations, and on current implementations on different systems, the lower-order bits are much less random than the higher-order bits'.
So, if he's on windows, he could fall into "much less (random)" category...
