specify an encoding explicitly, we used to treat it as being in database
encoding when we parsed it, but then perform a UTF-8 -> database encoding
conversion on it, which was completely bogus. It's now consistently treated as
UTF-8.