So Jesse (one of the founders of opscode (makers of chef)) used to work at amazon before he founded chef along with Adam. Adam had been working almost 2 years using puppet at other companies and it was pretty obvious he wasn't happy with the way puppet did things. I've heard that chef was closer to the method of thinking that amazon employed in their automated operations (which wasn't really exposed to the public). That being said, they're both fine solutions and are in use in a number of large enterprises (goog, etc.)...the key is to be automating your operations in some manner. Hell...use both and see which aspects are / are not important to you / your people.
"Chef's Data Bags are incredibly useful." Very true. Not having to deal with a DSL (which is a personal opinion) is the reason I chose chef over puppet. Plus the IRC channel seems to be more active.
Is the AWS opswork link broken for anyone else? I tried to view their pricing/sign up, but received 404's.
I like data driven stuff. Though there's still a DSL. This is a large reason why I made http://ansible.cc, as I wanted the actual description of the policy to be data as well.
True, I really appreciate the benefits of Heroku since it removes your concerns about server management altogether. OpsWorks may be more of a competition for services Engine Yard, though.
It'd pretty interesting to also know why Amazon went with chef first.