I've just had a press release from Salesforce.com about the next iteration of its software. It's all going to be based on cloud computing - you log onto it from a browser more or less anywhere and it's available to you wherever you are. Much as the new version of Apple's .Mac communications will be (rebranded as MobileMe) on the 11th of next month if you're an individual rather than a corporation wanting to use it. Much as Google's likely to do with Android on its phones.

I can't help thinking that this is actually a very old idea, rebranded as something new. When I was at college in the early eighties we didn't have browsers of course, but we did have a computer at school. This would link to a big computer at another school. You'd log on through the phone line and use the other school's big computer. Ours was called a terminal, theirs was called (I imagine) a mainframe.

This might start to sound vaguely familiar, albeit in a less sophisticated form, to people who're awed by this new 'cloud' concept. My own view is that apart from distributing the computing geographically this is old-fashioned computing all over again.

You are right that the technology borrows a lot from the past. In fact, you may call this a sort of "pattern" for shared system computing. However, I think enough different elements of the experience have changed significantly to call this "new":

- The business model is subscription or demand based. (Then again, perhaps the old timeshare models pioneered this...)

- The marketing model is focused on the consumer as much as the enterprise. (That is definitely new...)

- There is no fixed limit to the scale of systems running in the cloud, though most vendors limit long running or intensely large scale computations. (Perhaps the big difference here is you aren't getting a fixed time block, but rather a "cushion" in which you can grab a tremendous amount of compute capacity as needed then give it back.)

- A new architectural model in which computing crosses both institutional and political boundaries without the knowledge of the end user. You hint at this, but this is really where most of the challenges seem to be coming from these days.

Another way to look at this is we may finally be seeing the dream of an online market for commodity compute services come to pass. That definitely was not true for the mainframe era.