> if you already know how you will access your database, it's ok to tell the

database in

> advance - rather than having the database figure it out by itself after

500

> slow queries.

Already existing indexes should be respected until there is a better
estimation due to past system load. An index will be dropped if it is not
needed anymore and a new index is created if it speeds up the system in the
future. (I know this is unclear/vague!)

> Second, telling the database you don't need to keep the indexes up-to-date
> when importing a lot of data

Something like SET AUTOINDEX ACTIVE|INACTIVE could solve this.

> Third, UNIQUE indexes also enforce a constraint on the database.

I had more the idea of giving in a DDL file just the relations and
constraints (foreign key, primary key, unique, check, ...) and the rest is
up to the DBMS.

> However, I do think that a utility to examine a given workload and suggest
> indexes would be a useful thing to have in my toolkit.

So at least we agree at this point.



Alexander Bauer