Casting of 0 is always special in the language, since 0 is a synonym for NULL.

In the sibling language C++, the recommendation is to use 0 instead of NULL in the source code.

But in both languaes, the compiler is required to see a 0 as a NULL pointer even if the hardware would happen to use something else to represent NULL internally.

Anyway - I do not like indexing around a NULL pointer. A NULL pointer (which 0 is) may point to the start of memory range (which it does in most architectures, even if that address range may be empty and create read and/or write exceptions). But the standard doesn't guarantee that a NULL pointer points to any memory range at all. The architecture may have a hidden bit in all address registers to represent the address as valid or a NULL. Such a architecture would obviously not manage a type cast from a NULL pointer to another data type and then the addition of an offset. The magical "NULL" or "invalid" bit would still be set and any offset would still represent a NULL.