Copy data set ddname construction for JCL-allocated data sets
If any of the following conditions exist, the ddnames specified with the REGISTER, OUTCOPYDDN, and RECOVERYDDN options (or the defaults) are coded in the JCL as prefixes to the partition or data set numbers:
The object specification is for a specific data set or partition (DSNUM integer specified).
The object is partitioned and OUTCOPY BYPART is coded with the OPTIONS command.
The object is partitioned and the OUTCOPY installation option value is BYPART.
In the JCL, the ddnames must be in the form namenn or namennn, where name is the ddname specified with REGISTER, OUTCOPYDDN, or RECOVERYDDN and cannot be more than six characters for objects with less than 100 partitions or five characters for objects with greater than 99 partitions.
For a partitioned table space or index, nnn is the partition number and must be in the range 1 through 999.
For partitioned spaces with more than 999 partitions, you must use dynamic allocation. You can have 4096 dynamically allocated data sets.
For a nonpartitioned table space or index, nn is the number of a single data set and must be in the range 1 through 32 for table spaces, or 1 through 128 for indexes.
If the object specification is for all of the data sets of a nonpartitioned object (DSNUM ALL specified) or for all partitions of a partitioned object, and the OUTCOPY option is ASCODED, then the ddnames specified with the REGISTER, OUTCOPYDDN, and RECOVERYDDN options are used 'as is' in the JCL. When a request of this type is made, the copy made is for all of the data sets of the object. The ddname cannot be more than eight characters.
Using the specified ddnames in this way ensures that the output copy for each object or partition has a unique ddname. Output copies cannot share data sets, although output copies for multiple object partitions can share the same prefix because the partition number suffix makes each copy DD statement unique.