I know this is an old request, but if anybody is still interested, we do this by writing the Oracle data to a flat file with a predefined field-separated structure.
Then the file is copied to the mainfraime where it is read by a Natural program and broken up into the relevant Adabas tables/fields.