Subject: | |
From: | |
Reply To: | |
Date: | Thu, 9 Mar 2000 11:22:12 +0000 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
In article <[log in to unmask]>
, Steve Dirickson <[log in to unmask]> writes
>> not cause a problem. Beware that if you add the sort to all
>> three items
>> below, that your performance on puts is going to slow way
>> down if you have a
>> large number of detail dataset entries.
>
>Although, to keep things in perspective, it is probably fairly rare (as a
>percentage of all accesses) to add new records to such a set, and *really*
>rare to add a lot of them all at once. IOW, lots of DBPUTs, requiring the
>update of several sorted chains, is probably not something that these
>applications do all day every day.
>
>Speaking of the applications, Bill specifically noted that
> "The programs that use this Data set are supplied by a third party and I
>have no access to them."
>Things like switching to KSAM or doing retrievals on new fields added to the
>set are not options.
>
Although it would not be good practice without an explicit timestamp
field, it is just possible that the 'inaccessible 3rd-party application'
relies on the implied chronological sequence of addition of chain
entries to this dataset.
If so, and Bill adds a sort to them, that application may no longer work
as the designers intended.
--
Roy Brown 'COBOL programmers PERFORM it VARYING'.
Affirm Ltd
|
|
|