> This, to me, implies that .fromstring should accept bytearray (though
> probably not general buffer objects). In 2.x, I understand .fromstring 
> to initialize an array from machine bytes read into a string, but not
> .fromunicode. This is *not* a text method, and the result may vary for 2
> and 4-byte unicode builds. So I can see that 3.x needs
>  .fromstring(bytes) but not .fromunicode(string).


No, fromunicode is still needed.  It is used to fill type 'u' array with unicode wide characters.  In 3.x unicode type was renamed to str and str to bytes, so fromunicode should become fromstring (or maybe fromstr to avoid confusion) and fromstring should become frombytes.