General considerations

Variable-length records are processed with a record descriptor word (RDW) in positions 1-4, so the data starts in position 5. Fixed-length records are processed without an RDW, so the data starts in position 1. Control statement positions should be specified accordingly.

Your records can be EBCDIC, ASCII, Japanese, and data types you define yourself. To process Japanese data types with DFSORT, you can use the IBM Double Byte Character Set Ordering Support Program (DBCS Ordering), Licensed Program 5665-360, Release 2.0, or you can use locale processing with the appropriate locale.

Input and output data sets must be on devices that can be used with QSAM or VSAM.

Standard system data management rules apply to all data set processing. In particular:

Be aware that when using fixed standard record format for input data sets, the first short block is treated like an End of Volume. See z/OS DFSMS Using Data Sets for more details.
Be aware that, in some cases, if a DD statement specifies a data set for output that is extended to a second or subsequent volume, and another DD statement within the same step requests the same data set, only the records on the first volume will be read, and incorrect output will result.
Specifically, when a new output data set is allocated with a unit count and volume count greater than one, but specifies no volume serial numbers, one volume is allocated. If a second or succeeding DD statement within the same step requests the same data set, the same volume is allocated to it. If this job step extends the output data set to more volumes, this new volume information is not available to the second or succeeding DD statement.

Thus, you should not use different DDs for a data set to be used for output and then input in the same step, unless the data set cannot be extended to a second or subsequent volume, or is allocated with the guaranteed space attribute in the storage class. See z/OS MVS JCL Reference and z/OS DFSMS Using Data Sets for more details.

The maximum record length DFSORT can handle is subject to the following limitations:

Record length can never exceed the maximum record length you specify.
Variable-length records are limited to 32756 bytes.
VSAM variable-length records are limited to 32752 bytes.
Fixed-length records are limited to 32760 bytes.
Variable block-spanned records are limited to 32767 bytes.
For a tape work data set sort, the maximum record length is limited to 32752 bytes with NOEQUALS in effect and to 32748 bytes with EQUALS in effect.

Note: If AQ format is specified, or CH format is specified and the CHALT option is in effect, the maximum record length for variable-length records is 32767 bytes, less the length of the control fields.

The number of records that can be sorted using a given amount of storage is reduced by:

Processing control fields of different formats
Large numbers of control fields
Large numbers of intermediate data sets.

Providing an Extended Function Support program with an EFS01 routine can limit the record length that can be used when processing variable-length records.

The minimum block length for tape work data sets is 18 bytes; the minimum record length is 14 bytes.