Re: Hints in Practice

"Robert Klemme" <shortcutter_at_googlemail.com> wrote in message news:4l3u87F5edsU1_at_individual.net...
> Bob Jones wrote:
>> "Robert Klemme" <shortcutter_at_googlemail.com> wrote in message
>>> http://publib.boulder.ibm.com/infocenter/dzichelp/v2r2/index.jsp?topic=/com.ibm.db2.doc.admin/bjndmstr663.htm
>>>
>>> You see, it seems not to uncommon to have hints as a DB feature.
>>>
>>
>> Good links. I have worked with DB2 for years and have never even heard of
>> those hints. In fact, I have never seen a single line of code with hints.
>
> If you look at those pages you might discover that those hints are very
> much different from SQL Server and Oracle hints: hinting in DB2 seems to
> be restricted to manipulating statistics...
>

Yep, DB2 hints are implicit and they do not change the SQL.

>> The use of hints is discouraged. If you go to one of those DB2 classes,
>> they will tell you the same.
>
> That's basically what I stated in the first posting and asked for
> verification.
>

The topic has changed somewhat as it often does in this group.

>> I am not sure about SQL Server. Will they encourage the use of hints? I
>> doubt.
>
> I think, nobody really does. The only statement of you that attracted
> some fire was that about missing confidence. I rather see providing hints
> as a manifestation of realism: a modern DB engine has reached a level of
> complexity that it's realistic to assume that there are corner cases where
> automatic optimization does not yield optimal results; and the product has
> reached such a wide distribution that it's likely that someone stumbles
> into them. And so there is an additional handle for knowledgeable users
> to tweak. It's similar as with automatic transmission: most of the time
> you just drive in D but sometimes you need to manually shift gears - and
> you can do that with all automatic transmissions out there.
>

That's just another way of saying the same thing. Received on Wed Aug 23 2006 - 21:17:42 CDT