TERSE is an IBM archive file format that supports lossless compression. A TERSE file may contain a sequential data set, a partitioned data set (PDS), partitioned data set extended (PDSE), or a large format dataset (DSNTYPE=LARGE). Any record format (RECFM) is allowed as long as the record length is less than 32 K (64 K for RECFM=VBS). Records may contain printer control characters.^[1]

Terse files are compressed using a modification of Ziv, Lempel compression algorithm developed by Victor S. Miller and Mark Wegman at the Thomas J. Watson Research Center in Yorktown Heights, New York.^[2][3]

The Terse algorithm was proprietary to IBM; however, IBM has released an open source Java decompressor under the Apache 2 license.^[4] The compression/decompression program (called terse and unterse)—AMATERSE or TRSMAIN—is available from IBM for z/OS; the z/VM equivalents are the TERSE and DETERSE commands, for sequential datasets only. Versions for PC DOS, OS/2, AIX, Windows (2000, XP, 2003), Linux, and Mac OS/X are available online.^[5]

The following JCL can be used to invoke AMATERSE on z/OS (TRSMAIN uses INFILE and OUTFILE instead of SYSUT1 and SYSUT2):^[6][7]

//jobname  JOB ...                                            
//stepname EXEC PGM=AMATERSE,PARM=ppppp                       
//SYSPRINT DD SYSOUT=*
//SYSUT1   DD DISP=SHR,DSN=input.dataset            
//SYSUT2   DD DISP=(NEW,CATLG),DCB=ddd,DSN=output.dataset,  
//            SPACE=space_parameters                          
//SYSUT3   DD DISP=(NEW,DELETE),SPACE=space_parameters Optional temporary dataset

Terse can be used as a general-purpose compression/decompression tool. IBM also distributes downloadable Program temporary fixs (PTFs) as tersed datasets. Terse is also used by IBM customers to package diagnostic information such as z/OS dumps and traces, for transmission to IBM.