Stan Sieler wrote:
> I started to urge HP to provide a more useful "hook" for intercepting
> *all* IAMGE calls a few years ago, but the concept got lost in the
> work on jumbo datasets.
 
As an alternative, I asked DISC, Inc. (OMNIDEX) to provide "dynamic
timestamp indexes", e.g., a timestamp index that gets updated along with an
internal item index.
 
In this way, you can extract changed _item_values_ over a longer time period
than logging everything under the roof.  You can use Suprtool or other AdHoc
utilities or simply have a job sleep and wake-up to move the changed items
to destination, selected by date-time ranges quickly through TPI.  I would
think this feature would help sell more copies of TPI products on MPE.
 
The down side is that on index rebuild, you will loose the timestamps unless
they are stored/retrieved off to the side.  It is this step, I believe, that
dampened any enthusiasm of implementing this feature.
 
This lead to a discussion and question:  Is there any reliable method to
match "meta" data in an IMAGE detail _if_ the detail is re-organized?
Record numbers don't count.  That leaves either chronological chain path
order.  As last resort,  add a unique data item value (a timestamp!) placed
in the detail set by an application.  But the later is not feasible in most
cases because of application program changes.  Any ideas?
 
In this area of data warehousing - for example, trying to sync an Image
database to Oracle on a Unix system, is a bit messy.  Which is the crux of
this original post.
----------------------------------------------------------------
Eric J. Schubert                    Senior Data Base Analyst
Office of Information Technologies  Univ of Notre Dame, IN USA
(219) 631-7306                      http://www.nd.edu/~eschuber