>George, >you can parallel either format (csv or fixed length), >Fixed length is much faster as the system will not have to scan the data to find out the ',' to delimit each field. ...have you measured how much faster, or are you basing this on principle? I perform huge, high end loads from external tables and have achieved as much as 250,000/sec per CPU with variable length. -- //www.freelists.org/webpage/oracle-l