In release 3.64, the install script doesn't indicate the database's CHARSET and each field's COLLATION. Then input and output characters are expected to Latin1 (collation is latin1_swedish_ci). This doesn't matter as long as using ISO-8859-1(Latin1) characters.
But it matters to utilize the other characters such as Japanese (EUC-JP) because MySQL convert the query and its result character set between client-defined (PHP) character set and database-defined (MySQL) character set. In detail, please refer to MySQL manual.
With this conversion, all Japanese characters is mapped into unknown characters in Latin1 database. This setting is meaningless for non-latin languages.
If the database is set as UTF-8, the other characters than latin1 is stored correctly. There is a strong demand to set UTF-8 collation for the database.