RE: Optimizer change from 8i-9i

Hi Wolfgang

>while I agree with you that setting optimizer_features_enable to a back
>release is not the same as running the optimizer at that release level,
>this is actually not a good example in this context as
>complex_view_merging is set to false by =
optimizer_features_enable=3D8.1.7

The aim of my example was simply to show that even if you run the RBO =
different execution plans can be generated during the migration from 8i =
to 9i. Of course in this special case you could probably disable =
complex_view_merging in 9i to get the old behaviour. Nevertheless, with =
default configuration the RBO is different in the two versions. =
Therefore saying that the RBO hasn't changed since 7.3 is a tale!


Regards,
Chris

PS: My guess is that the core of the RBO itself has not changed (except, =
perhaps, to fix bugs). But, before a statement is send to it, some =
transformation can be applied (like complex view merging). Therefore =
another execution plan is generated because the RBO works on a different =
statement and not because the RBO itself has changed.
--
http://www.freelists.org/webpage/oracle-l

Other related posts: