Currently mwdumper always tells the database it's on a UTF-8 connection when
speaking directly to the DB over JDBC.

This should work correctly with the UTF-8 schema, but not with the default
backwards-compatible schema; the result will be conversion by the DB into actual
Latin-1.

Some sort of horrible double-conversion might be necessary, assuming that even
works...

See Also: T16379: mwdumper crashes on non-latin input characters
Version: unspecified
Severity: normal
OS: Linux
Platform: PC