Re: Measuring index usage

On Tue, 26 Sep 2000 09:05:06 +0100, Jonathan Lewis wrote:

>Two problems with that:
>
>First - the supplier will probably say you must not
>change the indexing.

I still can't see why there should be a problem here. The supplier is not the enemy, we share the same goals. Plus the supplier is not a devine authority with infinite wisdom (and neither am I). We should both appreciate the fact that we can learn things from each other. At least, that's the way we've always done business with them.

>Second - the supplier MIGHT let you drop an index,
>and not be aware that one of their developers has
>written a piece of SQL that (illegally) assumes
>that the data will return in a given order that happens
>to be dependent on a specific index.

That's a good argument for not changing any indexes on my own (a bit far-fetched, but you never know...).
It's _not_ a good argument for not bothering at all. You have to draw the line somewhere. If the supplier says it's ok to change an index and my own investigations show the same, I won't hesitate for one second to implement the change. Following your logic, it won't be possible to implement any change, no matter what the source is. Following your logic, it won't even be possible to _use_ the system. Changing statistics on tables could lead to different query plans, which could lead to different ordering of queries, which could lead to the problem you describe.

-- 
Jan

jantah_at_hot.mail.com
...and you know what to do with "hot.mail", right?