Two views on wireless: Is it ready to make waves or still adrift at sea?

by David Strom

I hate to rain on AT&T's wireless parade, but it won't be as easy a sell as Kendra suggests. The industry is once again going through another massive re-evaluation of market positioning (vertical and special-purpose markets today, horizontals maybe in a few years); corporate network managers end up doing most of the integration and mediation among multiple vendors to get the kind of systems they need; and the three harpies of price, packaging and performance are still a challenge to vendors and users alike. All of this means that wireless networks won't be a boom town for some time, if ever.

Sure, having all this mobility is sexy: I don't want to be tethered to my data any more than the next guy. It is fun to surf the net from my car or get an instant update on my inventory no matter where I am. But let's leave Buck Rogers for the movies: Instead of any dramatic event that will move wireless data products off the shelves and into the hands of eager users, this industry will see more slow and steady growth in the coming years. And wireless vendors run the risk of setting expectations beyond what they can deliver, which ultimately hurts their overall market.

Let's look at some of the limiting factors, shall we? First there is the positioning problem. When wireless products were first introduced over ten years ago, they were promoted as replacements for wired networks and wired applications. However, that promise soon fizzled, mainly because of the lack of any industry standards that could promote overall market growth and show signs of maturity. On top of this, products were bulky (the PC Card had just been invented but few laptops had them) and difficult to use.

About five years ago, vendors shifted focus to vertical markets and specialized applications, and seemed to find some success in shop-floor applications, vehicles (what I call "guys in trucks") and hospitals. All of these were just emerging niches and narrow ones at best.

About three years ago, RadioMail and other general-purpose office applications were supposed to ignite this industry: we were all going to become wireless email users. That never took off: the products were always lagging their support for Windows, radio modems only recently became available for PC Cards, and many laptops were radio-hostile.

The pendulum has swung back and forth so many times between vertical and horizontal applications that I've gotten dizzy trying to keep track -- but according to Kendra we are back in a vertical position this year. I think that is good, but the focus should be on connecting new employees to the corporate network in ways that wired networks either can't handle or are too expensive to do with laying new cable. Thus wireless makes the most sense for employees that move about too much and don't spend their day hanging around a fixed Ethernet drop, or for those that work in difficult-to-wire places.

Vertical markets are good at providing an entry point into corporations, but vendors need a good buzz or word of mouth to be able to expand beyond just the guys in trucks to the guys sitting at their desks. Unfortunately, word of mouth doesn't usually make for big growth spurts, getting back to my thesis that wireless won't be a three-day wonder but a longer-term investment.

If vendors are to succeed at making this crossover to the general purpose market, they'll have to have several things going for them today: First, they need a strong OEM business to find themselves in several distribution channels concurrently. Those that don't won't have enough channel bandwidth (pardon the pun) to make it. Xircom, Proxim, and perhaps AT&T and IBM fit this description.

Next, vendors will also need enough of a core business to generate the high level of cash and margins needed to play in the wireless arena. Wireless takes big pockets: not just in terms of building products, but paying engineers who understand radio devices and investing in manufacturing to build the devices. That tends to cut out many of the smaller players that currently are selling products.

Finally, they'll need lots of partners and lots of patience: the best wireless products will pull together several existing off-the-shelf solutions and integrate them together.

This last point bears further investigation. Indeed, it is a sore one for many corporate network managers who generally have gotten stuck with being the wireless integrator of last resort these days. Why? Because wireless vendors have typically been too narrow-minded in their pursuit of customers, and have failed to deliver a single-stop solution for corporate America.

Take a look at wireless email as an example: you have to purchase the radio modem, the email software, and the network access from three different vendors. Inside the box is firmware and software drivers from two more. And then you have to make sure that the network driver works with Windows 95, or worse that you have picked the right priced plan so your users don't spend a fortune in air time.

Contrast this with cell phone voice-only users, who can walk into a storefront in any medium-sized town in the country, and walk out with a working product with a single purchase: wireless data still has a long way to go on the integration front.

Sure, better integration is coming soon to your local multiplex: IBM, Ericsson, and Ardis all have promised more work in this area over the past few months. But I am tired of talking about this situation in the future tense: when will we be able to move beyond promises and deliver something that can shift the integration burden off the backs of network managers?

Another issue for network managers is the integration of their wireless and wired worlds. Right now, they have to have two separate sets of skills, two separate network management systems, and two separate application development paths: there is little that can be shared between the two. That isn't very appealing nor a recipe for success: for wireless data to succeed, it has to be knit very tightly into the known world of wired networks. There is some promise: CDPD, for example, is basically wireless IP: all well and good. But it will take more than just putting together a wireless transport to make a well integrated wireless application.

Let's move on to price, performance, and packaging: all three have been a thorn in the side of wireless growth for some time. First is packaging, or the actual physical dimensions of the various products themselves. Until relatively recently, the notion of a single one-piece design was still foreign to many products. Look at what is considered a "portable" cell phone today, and what was portable five years ago: the difference is several pounds of gear and lots more battery life. I rest my case.

Many wireless vendors still require the user to mount a bulky external antenna somewhere on their laptops (glue or velcro is thoughtfully provided). That's too bad, because no one wants to carry around lots of parts that they can lose while running for their connecting flight at O'Hare. Lugging around this stuff reduces ease of use, so a one-piece design makes the most sense. Until recently, however, one-piece designs were hampered by noisy radio emissions from laptops that were built without any awareness of their radio spectrum usage. For this market to take off, laptop vendors need to design machines from the start to minimize their radio frequency chatter.

Part of the packaging issue has to do with the lack of integration mentioned earlier: can I buy everything I need from one place? Not yet: just promises of what things could be in the future.

How about price? Wireless products still cost too much more than their unwired cousins. Now that I can get an Ethernet card or a 28.8 modem for less than $25 (I understand that soon they'll both come as prizes in cereal boxes), it is harder for wireless vendors to compete for this cheap connectivity. Many wireless LAN adapters cost more than $700 last year and this year they are still more than $300. Given this price premium, it is hard to justify a wireless network without some serious productivity gains, which traditionally have been difficult to quantify.

Finally, there is performance. Ideally, a wireless connection should be perceived by the user to have the same throughput and latency as a wired one. In practice, this is far from where it should be. Latency on RAM's network, for example, can be measured in many seconds: this tends to kill any perceptions of reasonable response time for interactive applications.

The problem is that wireless protocols have tons of overhead, and overhead tends to kill off throughput quicker than a car phone user can drive into the nearest tunnel. Most wireless LANs are lucky if they can pull a megabyte through the air per second: their effective throughput, when you add in all overhead, is still far less. With the coming of 100 megabit desktop connections, this makes for a ever-widening gap in performance.

So wireless data has lots of challenges to overcome: it is slow, it costs too much, and it is too hard to integrate into the general Ether of the office. Fix these basics and we'll see some action in this industry. But these changes are all difficult and will take some long-term efforts on the part of the vendors. Until then, you and I will get by using the most popular wireless application of all -- I'll call you from my cell phone.