> Seriously speaking, it seems that nobody else took time to summarize the
> alternatives like you. But there's another: a non updatable view based on
> another view based on another table. If the intermediate view needs to be
> non-updatable is a thing you could verify, because I've tried so many things
> in latest days that I barely remember where I started testing.

One of my applications has been running in the field for two
years now - we still havn't finished the test spec. Writing
the questions is a lot more difficult than writing the
answers.

I've been following this discussion - probably because I'm
aware that I don't use triggers enough in my applications. I
think I can see the questions here, and why in this case the
answer can be so difficult, but is this not a case for why
NULL is not a value - you don't know one.

In reality what is missing is a 'modified' flag in the
database which is set once when a row is first modified, and
then the count is only incremented when a modified flag is
changed. There a simple question? Does it have a complex
answer.

( Like the flushing problem, should this be on
ib-priorities? )

--
Lester Caine
-----------------------------
L.S.Caine Electronic Services