[Python-Dev] RE: Unicode character name hashing

[Tim Peters]

[MAL]
> Note that Unicode code which needs enough bits to hold a UCS4
> value should always use the Py_UCS4 type (see unicodeobject.h)...
> any changes needed to assure the 32-bit nature should then
> be made to the Py_UCS4 typedef.

That's fine.  The Py_UCS4 typedef should in turn be defined by direct
reference to a C9X-derived typedef from pyport.h:  the latter is where
32-bitness "should be" defined.  Then the only changes ever needed will be
to pyport.h (where people will *expect* such changes to get made).