On Tue, 8 Dec 2015 22:24:09 +0000, J O Skip Robinson wrote: >Professor Skip assumes that it will be done wrong--at least in execution. >Unless the design anticipates and properly handles all execution flubs, then >the design is wrong. What could go wrong? > The design should be made right.
>-- A faulty command is issued. If the update is echoed to the control file, >the component (JES2 in this thread) might fail to come up or at least work >properly at IPL time. > A faulty update is made to the control file, perhaps with the ISPF editor. ... I assume a well-managed system has controls to prevent this, perhaps a validation filter. The same filter should operate on "Save as Default". >-- A command is issued by an operator who may not even have READ access to >JES2 parms, yet the content is written into the control file. > ??? Doesn't RACF prevent updates, however indirectly, to a control file by an operator, programmer, admiistrator, CIO, or CEO who hasn't WRITE access? >-- Two or more commands are issued in quick succession--maybe by different >people. What gets echoed to the control file? > Two or more updates are made to the control file--maybe by different people using the ISPF editor. Which prevails? I'll assume locking of sufficient scope that this is as impossible as ISPF, for example, makes it. I've coded for SQL/DS (System R on CMS) and in my coding I needed to deal with situations in which two or more people made conflicting updates. One must be ROLLBACK and the other COMMITed. >If you take a poll of sysprogs on whether to implement this mechanism, I doubt >that many would be enthusiastic. > I admit I was somewhat in the blue sky. But the present situation where much of the configuration is by command rather than control file still strikes me as chaotic. An operator can easily issue a command that makes the system inoperative. The recovery is IPL, however unpleasant. How about tweaking the configuration by command, then after several hours (or weeks) of validation snapshoting the configuration to control file(s)? Don't PARMLIB members have some sort of two-digit numeric version suffix? Can the operator at IPL elect an earlier version? -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
