John,

> I'm less impressed after running some tests:

> 2) Inserting a new node involved updating all parent records - which
> is quite blocking. (those transactions had better be very quick
> indeed!)

I agree. I read the article a while back, but I remember the side effect of
an insert could be even worse. It's not only parents that must be updated,
but all what is on "right" of the inserted node. So no matter how deep the
tree, inserting near the "left" of the tree won't cost the same as inserting
near the "right".

> After testing, I'll stick with the old technique. With Firebird's
> fast stored procedure language, it performs very well.

John, would you be so kind as to post one of those SPs for the example ?
I'll need such code very soon, and I'll be happy to start with something
tested.

Best regards.

--
Jerome