Re: My 10 Commandments of Database Administration...

On Thu, 23 Sep 2004 08:38:23 -0500, Nelson, Allan <anelson_at_miswaco.com> wrote:
> Not so fast. The state for all records in fnd_login that have no
> corresponding record in v$session and v$process and end_date is null are
> both known and incorrect. Just because something was validly null at
> one point in time does not mean that it is true forever and this is the
> problem for leaving the nulls there when the user is not connected.

I'm with Mark on this one, subject to the rather important caveat that apps doesn't provide you with a way of determining when a session crashed (we don't run apps here).

Knowing that a session has ended does not imply that we know when it ended. I'd submit that NULL was still perfectly valid for this situation. A query that counts sessions with a null end date and regards them as current would of course be wrong (as Nuno states) but I submit that it isn't the value that is wrong but the interpretation of it as having some meaning other than UNKNOWN.

-- 
Niall Litchfield
Oracle DBA
http://www.niall.litchfield.dial.pipex.com
--
http://www.freelists.org/webpage/oracle-l