C compilers and non-ASCII systems

Well, it does have to be in the sense that it has to be possible to (somehow) refer to individual chars. It does not have to be char-addressable in the sense of there being any machine-code object which can be trivially indirected through to obtain a single C char. Indeed, there is no _requirement_ that C use whatever indirection the machine code level supports, though if there's a way to implement the C abstract machine semantics without too much hair, it should be used. I think there was once a C implementation for the Lisp Machine that represented pointers as <array,index> pairs - Lisp arrays, of course. Only if you never use the rest of the word for any purpose, because copying something by pointer casting to char * or unsigned char * and copying the chars has to copy the entire object. (The Lisp Machine implementation above presumably took advantage of the way that model of Lisp Machine allowed two arrays to overlay one another, sharing storage.) Yes and no. They have to share the same space in the sense that copying the chars making up an object copies the object. They do not have to share the same space in the sense that the conversion between (say) a pointer-to-int and a pointer-to-char pointing to the first char making up the int may be an arbitrarily complicated operation. (As a quality-of-implementation issue, it's probably better that it not be especially complicated, but that's not a conformance issue.) /~\ The ASCII Mouse \ / Ribbon Campaign X Against HTML mouse at rodents-montreal.org / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B