Multi-byte data should be treated as binary data. That is, DBMS and IIR utilities must be informed that they are operating on binary data so that they read the data using binary mode file I/O.

By default, IIR and Oracle utilities assume they are operating on text data, and will read and write files in text mode. If a character in the input file matches a newline (

CR/LF

) or End-of-File marker (

Ctrl-Z

), the input file will not be read correctly and records may be accidentally split and/or the whole input file may not be read. This is more likely to occur with

UTF-16/UCS-2

data.