You don't always have the luxury of knowing the language of the text you're processing. You may well encounter both English and Turkish words in the same text stream, and you don't always have the means to distinguish between them.
It's the same problem with Han unification. It's fine if you have 100% Japanese, Chinese or Korean and set your font accordingly. But it all breaks down when you try to mix two or more languages in the same stream, unless you have some external means of distinguishing between them.