I discovered this in the process of converting many S36 apps to native mode: Back in S/36 days, it was common to use ADDROUT sorts. These sorts produced files of three-byte records, containing the record address of the sorted record. It was more efficient than sorting the entire record. . . . . At least it used to be. When converting to native mode, I have observed a noticeable improvement in performance when ADDROUT sorts are changed to RECORD sorts. This is not scientific, but the improvement is observably dramatic, and it occurs 100% of the time. In order to effect this change, change the sort header from SORTA to SORTR, and include the record length. Add a data line to the bottom of the sort that contains the record length. Then change your program to include only the sorted file. The "E" spec, and the ADDROUT file may be dropped. Dave