Thanks David. Sometimes I wonder about that, why there are only 3 types of SCD, why don't we make a new type. Why only Kimball and Inmon method (and Barry Devlin), why don't we make a new method. Your question whether a data warehouse must be a dimensional has been in my mind for a long time now. I think dimensional is a better format to get the data out but normalised form is a better format for integration. I think real time ETL is really hot at the moment and it breaks the fundamental principles that a DW is a non-volatile store. Another trend is text analytics which developed very quickly in the last year or so which I think will be a hot potato in this 2007. What do you think?

Regards,

Vincent