> I'm using IB_Edit controls descendant, so they should be ok.

You mentioned IB_Date - it may be descended from IB_Edit but it adds a
lot. I dont know if it has any specific replication problems but it
is quite possible.

> But it's not this my main problem for this post :)

OK :-)

> Simpli put: take a dataset, a updatebar and a grid. Keep the dataset
> close. Then hit "+" and add a record, fill the fields, post it, then
> add one more, and so on. You will have, for istance, a grid with 10
> records. You can navigate the grid, edit fields, delete rows but the
> dataset is still Close! Is it "normal"? Is it "safe"?

It is intentional. It is a feature introduced a long time ago and I
have many users performing inserts directly from search mode. It
allows work to be done without wasting time retrieving records that
you dont want to see anyway. It is safe in my experience.

> If you have a program for enter orders, will you put the master
> dataset in insert state without opening it (since opening makes the
> select retrieve unwanted records from the server...) like I'm doing?
> If so, what about detail dataset? I had problems so I had to open
> the detail (never mind, since there are no detail records from the
> database for a new order ;)), and you?

I would have to check but I believe I have setup my application so
that opens are cascaded as a result of insert into the master. But
then all my datasets start in search mode anyway (from which I perform
the insert).

I have a fair amount of code backing up some forms so that all
controls appear consistently colored - so the user cannot actually
tell there are master-detail relationships.

hth

--
Geoff Worboys
Telesis Computing