Really good article but I must disagree with your comment that the ODS is not a data warehouse because it is in 3rd normal form, not in dimensional model.  Although storing data in a dimensional model is valid and may be the best model for supporting end-user reporting, it is not the end-all model for a data warehouse.  The only reason your ODS is not the definative data warehouse is because you said it can be updated.  This idea is not generally supported in the data warehouse... but as with anything there are exceptions to every rule i.e. Type-1 slowly changing dimensions.

I believe that different methods work for different environments and we must be flexible to real world environments yet frigid in staying with proven best practices... whether it's Kimball dimensional models, Inmons CIF and DW 2.0, or John Doe in Little Company USA's who made the data warehouse work within his real world limitations.  Business practices change, methodologies change... and who knows, in a couple of years we may have a new method for building data warehouse system... the Rainardi Method  

Anyway I digress, overall great article with a lot to think about and I'm looking forward to more.