Re: Poor PaaS, where is the love?
Actually, if you read PaaS forums (like Heroku, OpenShift, Azure), you'll see scores of customers who aren't able to fit their applications (often quite standard ones) into the arbitrary decisions that have been made by PaaS designers. (To be fair, both Java and .NET PaaSes tend to do better because of the VMs they run on, but even still, you can see complaints about implementation).
I don't think the column here asks the right question. The open question today is whether the PaaS model (Docker, OpenShift, Azure, Heroku, etc) will be the predominant one, or whether the configuration-management model (Chef, Puppet, Ansible, SaltStack, etc) will. The benefits to the latter are that you don't run into implementation issues; you can implement whatever you want. The benefits of the former are around simplicity.
Today, it's hard to argue that the PaaS side is winning. The overwhelming adoption of Chef by essentially everyone (AWS's OpsWorks, CMPs like RightScale, even Netflix OSS) combined with the relative silence from the startup community on using PaaS (who exactly uses it? RapGenius used to... single digits of well-known companies, right?) indicates to me that the future isn't PaaS merging into IaaS, but rather IaaS with a configuration management layer (which I guess you could call PaaS, but certainly isn't considered PaaS by any PaaS vendor I know).
The enterprise side is more in the PaaS world, but it's really hard to tell how much, given that they don't do much sharing. But even there, from anecdotal evidence, I believe config mgmt is winning.