The export file identifies the character encoding scheme used for the character data in the file. If that character set is any single-byte character set (for example, EBCDIC or USASCII7), and if the character set used by the target database is also a single-byte character set, then the data is automatically converted to the character encoding scheme specified for the user session during import, as specified by the NLS_LANG parameter. After the data has been converted to the session character set, it is then converted to the database character set.
During the conversion, any characters in the export file that have no equivalent in the target character set are replaced with a default character. (The default character is defined by the target character set.) To guarantee 100% conversion, the target character set should be a superset or equivalent of the source character set.
For multi-byte character sets, conversion is only performed if the length of the character string cannot change as a result of the conversion.
For more information, refer to the National Language Support section of the Oracle7 Server Reference.
This situation occurs because the eight-bit characters in the export file are converted to seven-bit characters via the client application. When sent to the database, the seven-bit characters are converted by the server into eight-bit characters. To avoid this situation, it is necessary to turn off one of these conversions. One possibility is to set NLS_LANG to the character set of the export file data.
When importing an Oracle version 6 export file with a character set different from that of the native operating system or the setting for NLS_LANG, you need to set the CHARSET import parameter to indicate the character set of the export file data. The CHARSET parameter is described .