The coming of the megacomputer

I'm as skeptical as the next man, but cloud is subtly different from that; it implies connecting to a service rather than a specific, named, individual server.

Cloud computing is important in that it is a new(-ish) paradigm that is a nice fit for certain tasks that require massive amounts of some resource to work (CPU, storage, something), and the amount of resources required lends itself to cloud-style management. For instance, backing up to the cloud is a decent idea, and on the cloud side, you don't want to be managing those backups machine-by-machine, you need something like cloud tools and cloud abilities for any one machine to go up in flames and have the services just keep going with no damage.

Cloud as magic-pixie-dust-that-make-everything-better is dumb. It's hardly any different from the client-server magic-pixie-dust-that-make-everything-better. It is called for in some circumstances, and in some circumstances it just gets in the way. Playing a 3D video game from the cloud is really stupid, for instance, despite having recently heard AMD has plans to try this. To make it work you have to stuff so much data down a relatively small line, network latencies and such will be awful (current 3D games have dealt with latency in your opponents but there will be no magic way to deal with input latency), and you don't get much because putting 3D hardware six inches from the TV just isn't that expensive. The price differential between a box capable of being a client to this cloud (I can't imagine it could be done for less than $150, it's going to be more complicated than a Roku box) and having a box capable of just rendering 3D directly using local 3D-rendering resources (XBox 360 Arcade, $200 right now) is just too small.

Serving applications from the cloud over a web browser is also brain-dead. You get the limitations of having to be in the browser (which is just never going to catch up to the power of using QT, as one example), and for what? All you really need to use a "wordprocessor in the cloud" is to take a conventional word processor (not limited by the browser) and hook up the "Save" and "Open" dialog boxes to the cloud, hardly any different that what things like KWrite can already do by hooking up to things over SSH or FTP or other such things.

So, cloud computing is a Next Big Thing, but it won't be the Next Big Thing. It's a hammer, but not everything is a nail, and for every incremental increase in the power of the cloud, my local processing power I can buy for $100 continues to go up too.