Does anyone know if the timestamp in a journal record is guaranteed to
be unique? We've got really fast processors these days.
If they are not, then timestamp + journal seq number would be unique.
That should cover the issue of the journal seq being reset when the
receiver is changed since it seems likely that a receiver change will
take more that a microsecond, or millisecond or whatever the smallest
resolution of the timestamp is.
Where do you keep track of the last values? Data areas are fast. I
don't like sub-stringing them either and probably would opt for two, one
for the timestamp and one for the seq number.
But you can't read data areas directly in SQL. A keyed file with a key,
a timestamp and a journal seq number would be easy to use. And you can
purge it with a very simple SQL statement; no reorg needed if you set it
to REUSEDLT(*YES). For bonus points use commitment control on either
the data area approach of the file approach.
On 9/5/2012 5:26 PM, Stone, Joel wrote:
So aren't I better off using a date/time window?