What you vaguely outline has already been standardised in UTS #10. The algorithm is both based on prevailing user expectations and also has shaped them since the wide-spread adoption of implementations.