It's been two months since Larry Ellison, in an onstage interview, treated the world to his latest, loudest takedown of cloud computing: "All it is is a computer attached to a network! What are you talking about? ... It's databases! It's operating systems and microprocessors and the Internet! And all of a sudden it's none of that -- it's the cloud! What are you talking about??" The audience at the Churchill Club in San Francisco roared its approval.

I used to think His Larryness would come around eventually. He has a long history of ridiculing the latest trend and, eventually, embracing it for marketing purposes. But Ellison has been singing the same mocking tune about cloud computing for awhile, and now others are joining the chorus. Just recently, both HP's Mark Hurd and IBM's Sam Palmisano have indicated their dislike of the term "cloud" and its inherent fuzziness.

[ InfoWorld's David Linthicum has his own interpretation of the rant by Oracle's CEO in Larry Ellison's cloud tirade: Why he's so scared. ]

And rank-and-file IT folks? I would say their reaction to cloud talk has gone beyond the usual snorting and eyerolling. Now they just tune out.

We all know why: Cloud computing is such a big tent that it has become a three-ring circus bursting with any technology or service, old or new, that touches the network. Is there anything revolutionary about software as a service? Or a host that lets you upload your virtual machines over the Internet and run them remotely?

If you buy InfoWorld's broad definition of cloud computing -- services sold on a subscription or pay-per-use basis over the Internet -- then the value or innovation offered by "the cloud" depends wholly on the services being offered. And so far, the focus has largely been on commodity services that duplicate common applications, development platforms, or server infrastructure.

But what about specialized Internet-based services that add new, advanced capabilities? Call them cloud computing or not, but such extensions make the value proposition a little more interesting: All the benefit of zero-footprint deployment, plus capabilities that would be difficult to obtain any other way.

I'm thinking of a recent conversation I had with Irfan Kahn, CTO of Sybase. He was talking about columnar databases -- an old idea now gaining adoption for analytical applications -- and map-reduce distributed processing techniques. Standard, SQL-based databases have well-known performance constraints. So why not just upload your SQL data to a specialized processing service in the cloud, have the provider convert it to "NoSQL" format, and run the job in a fraction of the time it would take you?