Interesting remarks form vv-kirov, though l'eomot's querie is not all incorrect.

The point is that the function char() will give an output based on the character set of the code page in the collation of the database (at least that is the behaviour vv-kirov's example shows).

So l'eomot's querie will only give the right answer if the collation has already a latin1 codepage base, because when running it on a cyrillic or japanese based collation its output can give totally other characters.

A "collate" function afterwards (in the "between" statement) cannot change the character itself, only its position in the order.