Nice blog post Jeff. Have you signed up for the Blogger Link program for BOB? Here’s a link to it: Automatic Blog Links Coming.
With regard to tables without primary keys not making much sense I would say that is very true for early implementations of data warehouses. However, many of our largest tables (both dim and fact) do not have declared primary key columns. That’s not to say that there aren’t available candidate columns available. In fact most of these tables used to have declared primary keys. We simply got rid of them when we switched to partitioned tables and local indexes. We felt confident enough in our ETL process that we chose to not create any primary key or unique indexes. We do still have surrogate keys but without any constraints.
I suspect that our implementation is an exception rather than the rule. If I were working with a client on a new data warehouse with a less than experienced crew I absolutely would have primary keys implemented all over the place.
eganjp 
(BOB member since 2007-09-12)