Intel recently abandoned its plans to add "soft access point" functionality to an upcoming PC chip set. Now, exactly why Intel would want to be in the soft-AP business I'm not sure: There doesn't appear to be any particular reason why a given chip set would be more amenable to soft-AP functionality than another, or even that such capability would help Intel sell more chips. Intel was criticized for this move by some; apparently the changing of one's mind is now disallowed even for those who are not presidential candidates. But there's really a much more interesting question afoot here.

A soft AP is a software implementation of a wireless-LAN access point that typically runs on a PC-turning a notebook into an AP for quick work group meetings, product testing or a fun afternoon of spoofing and hacking. Some, however, have postulated that soft APs running on PCs could replace the more traditional AP entirely in residential and even some enterprise settings, with every PC serving as an AP in what would truly be a dense deployment with no traditional WLAN infrastructure. While I could envision such working in a client mesh architecture (where the client is the AP), I have a real concern about the viability of this approach with respect to traditional APs.

Those traditional APs are appliances, which I define as microprocessor-based entities running firmware or software but not otherwise programmable by the end user. In short, they're nice, self-contained little worlds that (almost always) just work. They're cheap, they're reliable, but, to be fair, they're still boxes running code. So there's some appeal in replacing even this minimal hardware with really cheap software.

But like many of you, I'm sure, I cringe when I add new software (or otherwise make any changes at all) to my PC. Will it still work properly afterward? Will bizarre behavior show up at the worst possible moment? The PC has become a bloated monstrosity, largely thanks to an operating system that exists more to preserve Microsoft's monopoly than to serve the paying customer. Do I really want to put critical infrastructure into such a beast? Not today, I don't.

The real beauty of the appliance-based approach is that the functionality simply works the same way all the time. Simplicity breeds reliability. I believe the ghost of Thoreau will someday visit OS developers everywhere and remind them of their original mission. Maybe then we can consider replacing our appliances. Maybe.

Craig Mathias is principal of Farpoint Group (Ashland, Mass.).