On Wed, 30 May 2018 06:15:11 -0700, Charles Mills wrote: >Humor me -- I'm not a VSAM guy. Are there really people who are counting on >their COBOL programs blowing up if they fail to do the right juju on their new >VSAM file? > Not COBOL, but from a recent ply:
On Tue, 29 May 2018 21:29:25 -0400, Roger W Suhr wrote: > ... >I recently found out that some IBM utilities (IMS DBRC) actually expect a >cluster in "create mode", when switching to a new secondary RECON dataset. If >you provide an initialize cluster, the DBRC utility ABENDS and says the >dataset had been used before and no valid header record could be found. >So I doubt IBM will ever provide an option to do this in the DEFINE command. > Why do the VSAM faithful believe that if this were done it must be done wrong!? I see "option" as the operant word here. If an option were provided and the historic behavior were the default, there'd be no adverse consequence. My only use of DEFINE CLUSTER has been creating SMP/E CSIs. Then I (must?) REPRO GIMZPOOL to initialize the KSDS. I don't know whether this provides a header record. I don't know whether a fits-all analogue of GIMZPOOL could be provided in SAMPLIB. I don't know whether this necessarily creates a record with the risk of key conflict. But for SMP/E CSI it can be done in the same IDCAMS step, minimally burdensome. >There are programs available on the CBT Tape (CBTTAPE.ORG) that will >initialize a new cluster (intelligently). Personally, I like the INITKSDS >program. > ISVs, at least, dislike introducing dependencies on fourth-party products, particularly those that may be unsupported. -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
