Since my new dn_normalize() uses the unicode functions, the ucdata must be loaded before the first call to dn_normalize(). Currently dn_normalize() is called first time in read_config(), while the ucdata is loaded afterwards in schema_prep(). Stig