Input data sets can be either non-VSAM or VSAM, but not both. DFSORT accepts empty and null non-VSAM data sets, and DUMMY data sets, for merging (be sure to supply RECFM, LRECL and BLKSIZE). DFSORT also accepts empty VSAM data sets for merging provided VSAMEMT is in effect. For non-VSAM data sets, DFSORT examines the DS1LSTAR field in the format-1 DSCB to determine whether the data set is null or empty. If DS1LSTAR is zero, DFSORT treats the data set as null or empty. A null data set is one that has been newly created but never successfully closed. Null data sets cannot be processed successfully by the Conventional merge technique.

RECFM must be the same for all input data sets.

BLKSIZE can vary, but for a Conventional merge, SORTIN01 must specify the largest block size.

With fixed-length records, LRECL must be the same for all data sets. With variable-length records, LRECL can vary.

Data sets can be multivolume but not concatenated. If a SORTINnn data set is multivolume and null, DFSORT forces an EOF mark on the data set before use.

DFSORT may set what it considers to be appropriate values for missing attributes (RECFM, LRECL, BLKSIZE) of input data sets based on other attributes, or may terminate due to a missing attribute. If a missing attribute results in termination, or you don't want to use a missing attribute set by DFSORT, specify that attribute explicitly (for example, specify RECFM=VB).

See Data set notes and limitations for additional considerations.