Re: Freezing database statistics

Jeroen van den Broek wrote:
> Niall Litchfield schreef:
>
> [..]
> >
> > Oracle won't (or rather if it does its a bug) get the wrong result,
> > however it may decide upon a pretty bad execution plan because it
> > doesn't know what the spread of the data is.
> >
>
> If it _does_ get the wrong result, this would be a nice, modern version
> of the 30 year old "Database Halloween Problem".
>
> http://blogs.ittoolbox.com/database/technology/archives/what-is-the-halloween-problem-in-databases-12618
>
> --
> Jeroen

It would also be a bug relating to Oracle's fundamental read consistency model, which is supposed to be relative to the start of the transaction. The Halloween problem should give a mutating table error if you tried to do something funny to work around Oracle's model.

Wrong results based on histograms would be a fairly severe bug IMO.

jg

--
@home.com is bogus.
100 million websites and nothing to read.
http://news.netcraft.com/archives/2006/11/01/november_2006_web_server_survey.html