My original idea was to modify the first n bytes of the sdf file so that it looked like a dbf header and avoid reading the entire file using low level file I/O : The reccount would be determined by the sdf file size. the bytes over-written to accommodate the header would be appended to the end of the file. Fingers crossed this would negate the need to read the entire file.

This approach (to my thinking) will only work if the records are fixed length, and already contain th relevant deletion byte and no CRLF's.

That's it for today. The next time I connect I should be 'resident' in the U.S.

Seasons greeting to all.