In message <1165549775.331061.53010_at_16g2000cwy.googlegroups.com>,
baoqinye_at_gmail.com writes
>The architecture group of our company has told us to avoid using ORDER
>BY in our queries. Their reasons are:
>
>- In 10G, order by is incredibly inconsistent. They had an example with
>only 2 rows of data that was extremely slow. Even if it looks like it
>works alright, it's not guaranteed to, and we can't predict when
>it'll fail. That instability is one of the main reasons against it.
>
>- ORDER BY a primary key, especially when the primary key is not used
>in the where clause, has the worst performance and is definitely to be
>avoided.
>
>Is this true?
>
As others have said, this is probably complete rubbish. I suspect
inadequate memory allocated to pga etc, and inconsistent statistics.
Some evidence would be nice.
What are they proposing instead? Sorting in client programs? Data
arriving in random order?
--
Jim Smith
Ponder Stibbons Limited <http://oracleandting.blogspot.com/>
RSS <http://oracleandting.blogspot.com/atom.xml>
Received on Fri Dec 08 2006 - 04:09:32 CST
