On Tue, 15 Dec 2015 13:00:31 +0000, Vernooij, CP (ITOPT1) - KLM wrote: >-----Original Message----- >From: Peter Hunkeler >Sent: 15 December, 2015 13:53 > >From the DSS Guide: >Chapter 9. Managing space with DFSMSdss >Releasing unused space in data sets >... >Exclude data sets whose last block pointer in the data set VTOC entry is not >maintained in the VTOC by using the EXCLUDE keyword. This can occur if you use >an access method other than BSAM, QSAM, BPAM, or VSAM. DFSMSdss does not >release space for data sets whose last block pointer in the data set entry is >0. > So the designers committed the unforgivable blunder of failing to differentiate between "0" and "unknown". This has been a problem for me elsewhere, in various DB interfaces, where I want to select all rows in which a certain column is unset, and get, in addition, those in which it is 0, or "", or some other default value.
Rexx does it right: the SYMBOL function distinguishes a DROPped symbol from any possible explicitly assigned value. >There does not seem to be an option to override this restriction. > >-- > >Thanks, >that seems the explanation, as I guessed. >Still I wonder why these are not released, they have secondary space and >CA-DISK has no problem releasing them. Since they are SMS managed, they have a >decent EOF and there should be no danger of freeing space with valid data in >it because of an incorrect last block pointer. > The logic would need to modulate to distinguish an SMS-managed data set from non-SMS with a residual EOF not overwritten by BDAM updates. The likelihood of a wrong guess is small; the consequences devastating. And there's apt to be a user, somewhere, who deliberately (irresponsibly?) exploits the behavior and sets DS1LSTAR to 0 to suppress RELEASE. ISTR that at its advent HSM wouldn't migrate empty data sets, nor PDEes with no members. Perhaps a similar motivation. -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
