Joined: 23 Nov 2006 Posts: 19244 Location: Inside the Matrix
Hello,
As the message said, using keep on a new dataset may cause unexpected results.
Keep tells the system to neither catalog the dataset nor delete it - something that would rarely be desirable.
Specifying new,keep is prevented in many systems to reduce confusion and reduce unexpected results. One of the worst results of new,keep was that a given dataset might be on several dasd volumes.
One use of new,keep,keep was to write over external tapes that were not part of the local tape management system and were to be mailed somewhere for processing there.
Joined: 14 Mar 2007 Posts: 8797 Location: Welsh Wales
As Dick has said, DFSMS can change ,KEEP to ,CATLG which is better.
I usually set up a sweep job to delete uncatalogued datasets without backup once a week in my SMS environments, as some people uncatalogued their own datasets to stop other people copying their JCL examples.
Joined: 31 Oct 2006 Posts: 1042 Location: Richmond, Virginia
Actually, KEEPing without cataloguing would allow the same dataset name to be on several volumes, with the same or different contents, depending on what you wrote to them.
In the old days, users/depts might have their own removable packs, so cataloguing was not always important, so long as I knew what I wrote to my pack.
The analogy is that on a network, you can have a file name in your folder that might duplicate a filename I have in my folder, but there is no confusion, for us or the system; we know where we're looking and are authorized to look.
Nowadays, everything should be catalogued, but the IBM system does not require it, as that would be a major change in JCL requirements.