The biggest mistake Microsoft has made over the last decade is to confuse media with computing. (The illustration, by the way, is part of a satire.)

It's an easy mistake to make. Computing is now central to the production and distribution of all media. This is a big source of end-user and corporate computing demand.

There was great rejoicing in Redmond last week, I'm certain, when Comcast announced it would use Microsoft's DVR software.

So what could be wrong? Plenty.

Microsoft's emphasis on media is taking its eye off the ball, which is computing.

Computing isn't media, although media can use computing.

Computing is computing, and media is media.

That is, computing is not limited to media. Computing doesn't need media's inputs to happen. Inputs can come from anywhere. Computing shouldn't be dependent on media's outputs, either -- then computing just becomes consumer electronics.

I've said this before, and I will keep saying it until someone responds.

The evolution of the PC has ended. Microsoft is turning it into just another TV enhancement.

It’s time for a re-boot, one that starts with “exploding the PC.” That means abandoning the PC platform, and the painful upgrade cycle, embracing instead a new platform and paradigm, one in which components can be replaced easily while the rest keep going.

That platform, as I’ve said, starts with a modular, scalable wireless LAN. Upgrading such a system should be as easy as replacing the central radio, since 802.11g is backward-compatible with 802.11b. Storage and interfaces and processing can all be separate components, living where you want them to. So long as each component has an access point it can be easily changed-out.

In fact, an Always-On network can be upgraded without taking anything old off-line. There’s no reason why my wireless LAN couldn’t handle both my old machine and my new one. Just add new ports at the center.

Moreover if this network is constructed as a mesh, rather than as it is currently in residential gateways, around a central access point, then adding components should make it more robust. Cognitive radios in a mesh should be able to define, not just the scope of the network, but its limits, so that signals don’t run across your property line. This would be done by adjusting power outputs – just enough to reach the farthest reaches of the network, with orders for more-or-less power sent regularly across the link.

Then there’s the upgrade path. You just add. There’s no longer a need to subtract.

But the most important point is the one I keep harping on. When a wireless LAN is your platform, with the Internet available in the air, with PC processing available through the air, then the way is clear to developing a new generation of PC applications, applications that draw their data from the environment, as in Zigbee networks, and work automatically.

In an Always-On network data can come from anywhere – from your heart, from your soil, from your refrigerator, from the very air. (Wouldn’t you like to measure the level of pollution in your house and then buy just-enough scrubbing to control it?) It can be processed anywhere it’s convenient. You can interface with it where and how it’s best for you – from a touchscreen or using your voice.

Always-On is the Big Bang the PC industry needs to re-boot.