Re: [SPAM] 10gR2 Upgrade .. Watch out
- From: "Greg Norris" <spikey.mcmarbles@xxxxxxxxx>
- To: oracle-l <oracle-l@xxxxxxxxxxxxx>
- Date: Thu, 28 Dec 2006 07:46:48 -0600
As far as the default sort order changing, there are plenty of ways
for it to "break" without requiring an upgrade... just about anything
which changes the execution plan could potentially do so. Adding an
index, analyzing a table/schema, enabling parallel execution, changing
the hash_area_size setting, etc. Take a look at Tom Kyte's site
(http://asktom.oracle.com/) sometime. From what I've seen, he's been
able to provide a counterexample *every single time* someone claimed
that you can count on a particular sort order without use of an ORDER
BY clause (even within a specific release).
To put it bluntly, if you need data in a certain order then you *must*
use an ORDER BY clause... failure to do so is most certainly an
application bug. Upgrading the database seems a particularly
effective method of exposing such bugs, but it's *never* been the only
way.
--
"I'm too sexy for my code." - Awk Sed Fred.
--
//www.freelists.org/webpage/oracle-l
Other related posts:
