USA Today reports on secure organizations banning Wi-Fi: makes sense and I applaud it. If you don't have the resources to make sure that your Wi-Fi network is secure, and you're operating a going concern (whether high-security or not), you simply shouldn't run it. Developments in the coming weeks and months should make these current concerns moot as WEP gets its repair, VPNs become a given, and other methodologies and best practices for security continue to emerge. [via Alan Reiter]

Scholander's WISP thesis report from Sweden [PDF]: Joakim Scholander and two fellow students just finished their graduate thesis at Lund University's economics school in Sweden on wireless ISPs; this 94-page thesis analyzes the business behind these firms. (If you don't have a PDF reader, you can use Adobe's free PDF-to-HTML converter, which allows you to enter a PDF's URL and have the page displayed as text.)

Not NATty at AT&T Broadband: in the latest move to demonstrate their lack of technical understanding, AT&T Broadband is offering full-price networking gear from Linksys for their customers with NAT disabled. If you want to buy it at full list from AT&T, you have to pay $4.95 per additional machine on your network. The upside: these machines get real, routable IPs. The downside: very, very few consumer machines require static IPs and, in fact, would be better served with nonroutable addresses as the first stage in firewalling themselves. (The second stage would be a personal firewall package like ZoneAlarm for Windows or Intego NetBarrier for Mac.) The AT&T guy quoted in the link above said they typically track problems back to NAT, and they can't access machines for troubleshooting that are NAT'd. I'd agree with the first point; misconfigured NAT is probably a good reason why machines couldn't see the Net, but it's more likely DHCP configuration on the client machine that's the problem. As for the second, if I were an ISP of any variety, I'd have a simple program for Mac, Windows, and Linux that a user could run at their discretion that would diagnose network ills and, if it could reach a Net (via NAT or not), send some diagnostic information to me. That makes sense. No NAT? Not.

Intersil has decent fourth quarter: I don't pretend to a financial analyst, but it is nice to see the dominant Wi-Fi chipset maker producing good financial returns at the same time as it continues to push into smaller form factors, 802.11g chipset manufacturer, and other wireless ventures.

G, What about 3G?

Don't think I missed Verizon's launch into 3G space: I'm just not qualified to discuss the details yet. Verizon's launch is interesting, and Alan Reiter thinks it's real: maybe 40 to 50 Kbps is achievable. But one of the primary issues we'll find with cell data transmission (3G and otherwise) is that the pool of bandwidth that's shared among users of the same cell across frequencies winds up being pretty small.

In Wi-Fi, we may have 11 Mbps raw bandwidth to play with, of which individual users will burst to the speed of the upstream Net connection (or office wired LAN speed). Cell users, by contrast, will find themselves splitting much, much smaller pools in the (maximum) hundreds of Kbps. Just like I warned - a voice in the wilderness - that cable modem users would (and now are) ultimately be doomed by the very success of the pooled mechanism that gave them as much as 10 Mbps download speed, so, too, 3G and other technologies will wind up splitting bursty transmissions into tiny, queued pieces.

The answer, of course, is Palm's new approach, borrowed liberally from RIM Blackberry's success: continuous connections which handle most of the receipt and transmission as an ongoing task not as a wait-for-it operation. 3G handsets and data interface devices will serve their users best when they figure out how to spool information in and out: email is easy. But how about Web pages that you request and then a tone sounds when it's arrived? Other uses that will work just fine with tiny amounts of data including SMS and directory service or similar low-impact information requests.

Remember, though, that all of this will be severely limited by the metered charges and people's initial sticker shocks on their first bills - even the early adopters. If I'm paying per byte, I have to think very carefully about what I send and receive. Wi-Fi hot spot installations so far has largely been about time (minutes, 24-hour session, per month), not about bytes, because the upstream cost is a fixed DSL, T1, or other provisioned circuit.

What it boils down to is that number crunchers, marketers, salespeople, and executives like to think about units in their own industries: K is what the cell telcos are paying for. Consumers and businesses, perversely, don't think about cell minutes or data bytes or the rest: they think in chunks. A conversation. A file, An email. How problematic for us to learn to think the way companies think in order to use their technology.

I defer to Alan Reiter and his apparently infinite knowledge of this field for additional commentary. Read him. Subscribe to him. Possibly worship him.