> 
> Hmm?  I think it's an excellent way to smoke out those hidden
> assumptions about word size.
> 
Those assumptions are not especially well hidden, things like:
	char c[4];
	int32_t b;

	b = c[0] << 24 | c[1] << 16 | c[2] << 8 | c[3];

are not uncommon in any program. This example may be handled either by
using int_least32_t and (depending on situation) shift left by CHAR_BIT.
Another solution would be to have a compiler-specific type that is 32 bit
and use that to get an int32_t type.

-- Ragge