Subject [IBO] Re: How to maintain coherence between 2 databases
Author henry FRANQUET
first, thanks for your answers

--- In, Helen Borrie <helebor@t...> wrote:
> So--the "best strategy with IBObjects" would be to set up an
> application-independent replication service using the IBO simple
> replication components.

Some explanation about my application, needs and fears!
1) It's the same program which is running either with database
server or with local database. In case of local database (on laptop)
there is no DBA, just one end user. The application have a tight
control over the local database.

2) It seems hard to define triggers to create a log of inserted,
updated or deleted records as I have no control over the metadata
(the application allows the DBA to add tables and fields)
I have read IBO replication module help file and I should have read
it earlier : it looks fine! special interest on data transformation
for another usage! but for my purpose I still have some questions :
a) what about generators ? I have tables wich primary key is a
generator and there are referential constraints on that primary key.
When updatating server database I have to change all the fields
using the local value of the generator with the server value. Worse,
for statistics purpose, in a few case, I have the generator's value
without having referential constraint (I need only to know that the
owner of the record is different and I don't matter who it is, so I
have no referential constraint in order to delete the owner when it
is of no more use)
b) avoiding blob transmission when there are no change is a good
thing, but is it possible to have the same thing withother kind of
field ?
c) for replicating only some fields from a table, I have to
create a select statement as said in RPL Online help. But I can only
create such a statement in a delphi application, for the statement
must be built (as I've said earlier, client DBA may add fields)
d) Could I expect a good performance with a 56K modem (in
comparaison with file transfert)
e) last, as server updating is automatised with replication, what
about dual updating ? I mean on a local database a field has been
updated and the same field has been updated with another value on
the server side. I would like the end-user to say which value is the
good one (as it should seldom occurs)

3) In order to avoid problem with different generator's values
between server database and local database, I plan to replace local
database with an uptodate version of the base after updating server
database and also because there is much more modifications to apply
in that way

4) I have had a bad experience with Oracle replication between a
central server and about one hundred agencies (a gigantic slowdown !)
and I'm very aware of performance for this application (not to
slowdown the server during replication)

thanks in advance