[Marc-Andre]
> BTW: I wonder how long it's going to take before
> someone figures out that our merge sort based
> list.sort() is vulnerable as well... its worst-
> case performance is O(n log n), making attacks
> somewhat harder.

I wouldn't worry about that, because nobody could stir up anguish
about it by writing a paper ;-)

1. O(n log n) is enormously more forgiving than O(n**2).

2. An attacker need not be clever at all:  O(n log n) is not only
sort()'s worst case, it's also its _expected_ case when fed randomly
ordered data.

3. It's provable that no comparison-based sorting algorithm can have
better worst-case asymptotic behavior when fed randomly ordered data.

So if anyone whines about this, tell 'em to go do something useful instead :-)