The current support of unicode with 1 or 2-bytes characters all from the same codepage is good.
But I would appreciate getting a real UTF8 support (multibyte chars) that would allow to mix in the same php file chars from several codepages, eg Russian + Greek.