Clouds and Key-Value Databases

I was watching an interesting video on Computing Strategy in the Cloud Era, by Lew Moorman, CSO of Rackspace, speaking at the Glue conference.

In the Questions section at the end he talked about lock-in and mentioned the new generation databases such as SimpleDB as representing lock-in because they aren’t standardised.

I did some quick skimming and found an interesting debunking article on these key-value databases.

The enthusiastic response to these databases makes me wonder, apart from the obvious Google must be right attitude, if a sloppy view of the world as a soup of key-value pairs is actually an inherently comfortable way for many people to think about data? Is it appealing because this is a level of complexity we can think with more naturally?