Actually since each batch job is working with a different detail dataset the
program does the following:

1. Lock old dataset
2. Lock new dataset.
3. Read all records in old datase and write to new dataset.
4. Unlock old dataset.
5. Unlock new dataset.

This stategy was adopted to save some time since we only have to lock at the
beginning and unlock at the end.  We did not think this would cause a
problem since the detail datasets are "stand alone", ie: they do not have
links to any master or detail datasets.

-----Original Message-----
From: Leonard Berkowitz [mailto:[log in to unmask]]
Sent: Tuesday, March 30, 1999 2:11 PM
To: Carl McNamee
Subject: Re: [HP3000-L] Performance/image question




You are locking the datasets around each transaction, no? That's now how it
seems
from the way you wrote.

You should be doing:

Lock old
Get 1 old record
Unlock old
Do y our conversion stuff
Lock new
Put 1 converted record
Unlock new
back to the top.
========================
Leonard S. Berkowitz
mailto:[log in to unmask]
phone: (617) 972-9400 ext. 3250
fax:   (617) 923-5555

(leaving work now; will return on Monday)