I have a Swedish forum there some of the users have Swedish characters in their user names. I have now noticed that if two users have user names that looks alike but are different, they will not be able to login.
One user has the user name "daniel" and the other user has the user name "däniel".
I can confirm this misbehaviour in one of
<ins>I have access to
</ins>. The reason is the use of the collation
utf8_general*. A blog entry (in german language) offers a solution, the use of the collation
utf8_bin. But this has the same issues like your unique index and has to be implemented equally careful.
In the case with the user name, it is a problem. But in general I think this is a very useful feature. For example, a danish user may search for a word with the letter Ø then the word contains the Swedish letter Ö.