I'm the kind of developer that likes things under control by writing some stuff myself and keeping it as optimized as possible. This is especially true for database scripts. Sure we have these ORM tools these days (Entity Framework, NHibernate and such) but the calls that end up on the database side are many times unoptimized. Why? They have to work for all cases so sacrifices are made and performance usually suffers. That's why I like micro ORMs that make it super simple to translate database data to code (via object mapers) but still let you write optimized queries suited completely to your database.
The way that I upgrade my database requires me to backup existing data, drop the whole model, recreate it and restore data back to appropriate tables. Especially the backup strategy is problematic as it uses select into statements. Windows Azure requires all tables to have clustered indices prior to inserting data which makes it impossible to use select into statement as it has no control over clustered indices. How can we then automate backup scripts without creating those backup tables manually? With a bit of work, everything can be done.
