Vesa's point reminds me: is there any notion of the best way to have arrays and vectors with more than 2^31's elements (on 64-bit machines)? There are definitely times when I would like it even (I think) when an int is only 32 bits.