GPO (8/7/2013)
Wouldn't it be good if they created a licensing structure that forced db designers to normalise their OLTP databases properly. I'm heartily sick of people going on about how database XYZ must be important because it occupies 150GB on the disk, and we need all this RAM etc etc. Then you dig into it, and it's a mess of unnecessary clustered PK guids, ints that should be tinyints, nvarchars that should be varchars, redundant columns, strings that should be lookup values and on and on it goes. The smaller the db the cheaper the license and the world will slowly become a better place....
You bring up some good points, and others which I'm a bit confused about. Good database design would drive people to use the proper datatypes where appropriate. What I think a number of people forget though is that normalization is analysis, not design. As Steve mentioned there are some cases where specific denormalizations will improve performance and require fewer RAM and CPU resources, e.g. a current status of an item in the item's record itself instead of having to search through the history of the item to find the status.
As for licensing itself, Andrew's analogy of what you describe to Cloud data services makes me think we're talking apples and oranges here. If businesses are paying for the extra CPU and RAM hardware for in house servers and also paying extra for the software to use that hardware, it's a double hit to their budget for something that the software company didn't really do anything to earn that extra money. SQL Server would be coded the same for a 4 core system with 4 Gig of RAM as it would for a 16 core system with 64 Gig. In the cloud you're really paying only once for the "service" so it makes more sense in that model to price on size.
What I'd like to see is a more a-la-carte pricing by feature approach so that every SMB doesn't esentially have to pay the Enterprise edition penalty for just a handfull of features they will use that aren't in the Standard edition. I highly doubt MS will ever go that route since they like big software bundles, even though a feature based approach would probably help them better determine how people are using their software and what parts of the system are worth MS putting more research and development into.