Access Isn't Always The Killer Application

A great number of usage models fall under the "wireless" umbrella. For some users, access to a network infrastructure is the killer application. For others, the "killer app" is the capability to communicate independently between devices within a wireless personal-area network (WPAN). Usage models for WPANs include device-to-device data exchange, real-time audio, device control, and network access.

The mobile devices that implement these WPAN usage models are constrained by a number of factors including power consumption, bandwidth, and security. Because of these factors, the devices favor independence over global network connectivity. Some technologies manage to meet these constraints while accommodating the aforementioned usage models. Among those technologies are Bluetooth and IrDA.

At its most fundamental level, wireless data communication consists of computing devices (large or small) that communicate in some fashion without the aid of wires. Little else can be stated that will apply consistently across all types of wireless data communications. The specific nature of the communication between devices depends upon the usage models for which these devices were engineered.

Over time, a number of usage models for wireless data have come into being. For the individuals who depend upon wireless-data technologies, those usage models have come to define "wireless." For example, some people think that "wireless" means cellular-telephone communications. To others, "wireless" means Internet connections for mobile devices-whether those devices are associated with a cellular phone or not.

To quote Imrich Chlamtac, "This myriad of new devices spans the technological and budgetary spectrums, making it harder to attach the definition of mobile computing to any single device or application. However, there is one underlying similarity among all these devices and applications that allows us to simply define a mobile computer as a computing device, which can communicate through a wireless channel."¹ Hence, no particular usage model can claim a monopoly on what is meant by "wireless."

Wireless data communications can be analyzed along several spectrums:

1. The size and capacity of the device
2. The nature of the transmission medium
3. The nature of the information that's being communicated

It's common for the nature of the transmission medium to be somewhat dependent on the device's capacity. It also will be affected by the nature of the data communication in which it engages.

For wireless data communication, one of the most compelling usage models is to extend the Internet's reach. Television and radio advertisements speak of sending messages from device to device and making Internet purchases with a handheld phone or PDA. These days, efforts to enable mobile computing abound. This usage model has captured a great deal of attention in the mobile-computing community. In fact, it led to the now-common mantra (originating from the BARWAN project at Berkeley): "Access is the killer app."² I agree strongly with this statement as long as it is qualified with the addendum, "…for certain usage models."

If we presume that access to an infrastructure is the killer app, such an assumption can slant our perspective. It also can lead us to ignore other critical usage models, which are of great value-even if they have little or nothing to do with extending the reach of the Internet or the mobile office. Indeed, after providing the definition of mobile computing that was quoted above, the authors allow the "access" usage model to dominate their view of the wireless landscape: "A fundamental requirement of today's computer is network connectivity. This implies, mostly, radio-frequency wireless communications for mobile devices." Thus, they unnecessarily limit "wireless" to both a usage model and a transmission medium. In doing so, they limit the discussion of other significant and powerful usage models. They also omit relevant wireless transmission media that should be valued in the dialogue concerning mobile computing.

The notion of wireless personal-area networks implies a number of possible usage models. Most of those models don't depend upon access to a communication infrastructure. For example, imagine a pair of headphones that wirelessly connects to a home stereo, portable CD player, laptop, or television with very little user intervention. Because of the limited range of these headphones, a great deal of simplicity is introduced in the device hardware and communication protocols. This simplicity permits the headphones to more accurately and effectively meet a user's needs without a great deal of configuration.

Now imagine that those headphones require an IP address. They also demand some means of authenticating and connecting to an audio device via a wireless network within the home. As a result, the complexity, cost, and power consumption of the headphones will rise. To allow the headphones to function, greater configuration and infrastructure will be required. In addition, the ability for headphones to connect to consumer devices will always depend upon a network infrastructure that can't be guaranteed to exist everywhere that the headphones and CD player might wander.

In this situation, access to the broader network infrastructure is not the killer application. Rather, the killer app is a point-to-point connection between the headphones and CD player. This connection doesn't depend upon any other network infrastructure. Similarly, handheld personal digital assistants (PDAs) like the Palm, Handspring Visor, and Pocket PC can exchange information via IrDA infrared connections in a dynamic, ad-hoc, point-to-point fashion. They don't need to depend upon any other enabling wireless infrastructure.

Whether or not the "beam-me-your-business-card" or "mobile-headphone" usage models are the end-all, be-all of wireless communications, they require solutions that are suited to the following: the capabilities of the devices, the type of data that they need to communicate, and the nature of that data exchange. One size need not fit all. The fact that one user finds value in beaming a business card in a dynamic, ad-hoc, point-to-point fashion over infrared should in no way be seen as a threat to other usage models. These other scenarios might find users communicating from a cell phone over several miles via radio to a cellular tower. Or they could be communicating from the tower to the wired telecommunications infrastructure. Both scenarios have a place in wireless communications. WPANs provide an umbrella for wireless usage models that don't require access to a network infrastructure.

As Brewer points out, the notion of access as a killer app was largely a response to the prevailing trend in 1994 of handheld devices as communication islands. Each device would seek its own killer app as an information organizer.² In 1996, Kleinrock identified the disconnected state as the "usual" one rather than an "exceptional" one.³ Wireless islands are data networks that remain largely separate from telecommunication networks and handheld devices. They communicate with very few (if any) other devices (FIG. 1).

In 1998, Brewer and Stemm proposed a system in which individual mobile devices would no longer be isolated islands in the great wireless sea.^2,4 Instead, they would have the same access to the infrastructure as any other device on the network. In this communication system, handheld devices have gained access to the same network infrastructure as desktops and other devices (FIG. 2). If any two devices wish to communicate directly, however, they must do it via the infrastructure. This requirement is problematic in places where the infrastructure doesn't exist or one or more of the devices lacks the capacity to access the network directly. (Notice that in Figure 2, the network and telecommunication infrastructures are beginning to interact. They are permitting such uses as receiving e-mail via one's cell phone or carrying on a telephone conversation over the Internet.)

The WPAN environment allows devices to connect directly to one another without depending on a network infrastructure for peer-to-peer interactions. Note that this proposed environment doesn't preclude any given device from also making use of a network infrastructure when it has the capacity and need to do so (FIG. 3). In the future evolution of the wireless communication systems, handheld devices can communicate with each other as well as with the network and telecommunication infrastructures.

This systems view of wireless connectivity prompts many questions. For example, what usage models are best suited to WPANs? At the highest level, the usage model is largely unconcerned with technology per se. Instead, it describes the nature of the information that's being exchanged. It also details the manner in which this exchange takes place from a user's perspective. For example, a user may point a PDA at a printer and tap the print function. The data on the PDA is then printed. Perhaps another user turns on a pair of headphones and walks near a laptop on which an MP3 player is ready to launch. When the music starts playing, the sound is heard on the headphones. To understand the usage model in which they participate, the users don't need to understand the technical details associated with the data exchange.⁵

Specific usage models are usually best suited to particular technologies. Infrared, for example, tends to be ideal for dynamic point-to-point object exchange. It fits that role because of its natural spatial separation, limited range, and high bandwidth. In contrast, the longer range of RF makes it better suited to usage models that don't require high bandwidth, but must not be severely limited by either direction or distance.⁶ The headphone example is a good usage model for RF.

WPANs are particularly well suited to a number of usage models. One such model involves device-to-device data exchanges (FIG. 4). In this important class of usage models, data is exchanged in a point-to-point (or point-to-multipoint) fashion between two or more devices. This usage model ranges from simple object exchange to more elaborate synchronization and device browsing.

One of the simplest usage models for device-to-device data exchange is the object push. In this model, one device must be in a receiving state. The user of the other device issues an explicit command to push a particular object. This usage model is typical in printing and the exchange of data objects between PDAs.

A usage model that's related to object push is object pull. Here, one device makes specific requests of the other. In its simplest form, object pull may involve extracting a default object from a remote device. Imagine a meter reader driving up to a house and connecting wirelessly with the water meter. The water meter may have a wireless interface and a default object, which contains the information desired by the meter reader. The meter reader can send a short message that he or she desires the default object from the water meter. In response, the water meter can transmit that object back to the meter reader's handheld data-collection device.

In more elaborate situations, object pull may involve complex objects and structured interactions that transcend the notion of a default object. In order to navigate around a remote device, some kind of remote browsing capability needs to be implemented. In this situation, one device exposes a remote file system to the other device. Depending on the manner in which remote browsing is implemented, it may be made to function as a form of peer-to-peer networking. Here, the full file and folder structure of the remote device is exposed to another device as part of its view of the networked world.

To implement synchronization, these various capabilities can be employed with additional sophistication. In its simplest form, synchronization may involve pushing or pulling objects to establish a baseline backup of data objects. In a more sophisticated form, it could involve elaborate schemes that are designed to leave either of the devices reflecting information that was gleaned from the other.

Another type of WPAN usage model involves the transmission of real-time audio data. Real-time audio capabilities are common in many consumer-electronics devices (televisions, stereos, music players, etc.). Increasingly, these capabilities also are found in mobile computers like laptops or PDAs. Typically, RF is the preferred technology for these usage models. For the user, it allows a full range of motion within the spatial range of the device to which the user connects.

One common use for real-time audio is in wireless microphones. Usually, these microphones connect to a base station that was built specifically for the microphone. One of the objectives of WPANs is that these communication protocols become standardized around usage models. Devices will then be able to interoperate on a large scale with devices of a similar type-irrespective of manufacturer.

Wireless communication between speakers or headphones and sound systems is similar to this model. Wireless speakers would eliminate the need to run wires to remote locations as the consumer positions speakers around a house. The headphones provide mobility for the user as well as flexibility (FIG. 5). After all, the wireless headphones could be used to communicate with a number of devices ranging from the home stereo to the laptop and portable CD player.

Another interesting application that relates to wireless speakers involves communication between a cellular phone and an in-dash automobile speakerphone. Most cell-phone manufacturers sell hands-free phone kits that include a headset with both microphone and ear piece. This headset could certainly be made wireless, thereby introducing some improvement in the current usage model. In addition, the same technology that connects the phone to the headset could connect the phone to an in-dash microphone and speaker in an automobile. Thus, a user could utilize a cell phone in the car but dial through a steering-wheel-based touchpad. He or she could then communicate via an in-dash speakerphone.

Finally, real-time audio usage models may involve various forms of cordless telephony. Examples include a home intercom system, a cordless phone that connects to a wired base station, or a point-to-point capability for cell phones. For instance, Bluetooth includes a usage model that's called the three-in-one phone. It includes three usage models for a single phone. The first usage model is traditional cellular capability. For local calls, the cell phone can function as a cordless handset and connect to the wired base station (thus saving cellular connection minutes). Lastly, two phones can make a direct, point-to-point connection. They would then function as a sort of mobile ad-hoc intercom system.

A number of usage models also involve the control of peripheral devices. The most common version of this model is the television remote control. Typically, the remote control utilizes unidirectional diffuse infrared as its transmission medium. Because these devices don't require communication in both directions, diffuse infrared may provide an excellent solution for them. It can flood an immediate area with a signal while not interfering with devices outside the room. For peripheral devices that require communication at distances beyond 1 m and in both directions, however, RF solutions may be better suited. For a peripheral device, the longer distance and omni-directional nature of RF allows greater freedom of orientation.

Any peripheral device that's traditionally connected via a wire is a candidate for a wireless solution (whether RF or IR). Keyboards, mice, and other peripherals offer great value when they're more mobile. Laptop docking stations can be similarly liberated from the need to fit the external bus connection from the laptop into the docking station. Any arbitrary consumer-electronics device can be controlled wirelessly. Although this control could extend to toasters, alarm clocks, microwave ovens, and home-security systems, it may be difficult to see the exact value in controlling these things wirelessly.

Although access to wired infrastructures is not a prerequisite for WPANs, it isn't forbidden or prohibited either. For many devices, a connection to the Internet is a very valuable usage model. The access-point usage model will then provide a mechanism for connecting wirelessly to a device that functions as a bridge to the rest of the world. Cellular towers for cellular communications play this precise role. The towers communicate wirelessly with cellular phones while providing wired access to the telecommunications infrastructure. Any wired infrastructure can be reached via such an access point. For example, a network access point can provide connectivity to a local-area network. Presumably, it would provide Internet access as well.

A WPAN technology also may act as a bridge between a mobile device and another infrastructure. For example, a cell phone that's equipped with either RF or IR capability could establish a connection with a similarly equipped laptop. It would permit the phone to function as a modem for the laptop. In this case, the connection between the laptop and the phone is wireless-but not the sort of wireless that is used for the cellular connection.

In every situation in which data can be transmitted via a wire, that wire can be removed and replaced with a wireless solution. Some of these wires, such as phone wires, can be replaced with cellular technology. Other wires-like network cables-can be replaced with wireless-Ethernet technology. Many other wires go to speakers, headsets, PDA cradles, mice, keyboards, and a thousand other devices. Those wires also need to be eliminated.

When it comes to wireless solutions, one size need not fit all. Each wired technology fills different needs. Hence, the wireless replacements will differ in many respects. Cellular and wireless-LAN technologies do fill a great need. For many usage models, however, those two wireless technologies are not necessary-nor sometimes viable-solutions. These short-range, wire-replacement technologies are part of the broad set of WPANs.

Now that the various WPAN usage models are clear, it's time to examine the factors that favor independence connectivity over traditional network infrastructures. Many discussions of mobility make a tacit assumption about a largely homogeneous transport. In some cases, discussions involve heterogeneous transports. But they still assume a consistent usage model (namely mobile Internet access). They also assume a consistent underlying transport protocol (IP).

In "Vertical Handoffs in Wireless Overlay Networks," the topologies of the overlay networks are presumed to be different.⁴ But the devices that are participating in these vertical handoffs are assumed to be mobile IP nodes in a networked environment. For that environment, access to the infrastructure is a key function. The type of wireless transport may vary, but each device must be capable of supporting the various protocol layers that are required to implement TCP/IP. In addition, each device must be capable of communicating at varying distances via different physical media.

In reality, consumer devices tend to be built with capabilities that are sufficient to implement the usage models that are required by users in a particular target market. Makers of handheld devices tend to be concerned with three factors: size, cost, and power consumption. Additional computing power generally requires more expensive processors, a larger product footprint, and greater power consumption. Hence, it's not common for manufacturers to overbuild handheld devices. In other words, the notion of every handheld device being an IP node on the network is simply not tenable. But devices that are part of WPANs are only constrained to provide sufficient capability to implement their specific usage models.

When it comes to mobile devices, some issues are usually constrained. One such issue is power consumption. Two basic aspects tend to influence the power consumption of mobile wireless devices. The first one involves the capability of the processor. In general, the power consumption is greater if the processor is faster and more powerful. Because users tend to be dissatisfied with devices that run out of power quickly, manufacturers are extremely concerned with battery life.

A second critical issue for power consumption in wireless devices involves the chosen wireless topology. As a general rule, RF consumes more power than infrared. If the distance is longer for either technology, the power consumption will be greater. As a consequence, mobile devices tend to be fitted with a wireless technology that best suits the desired usage model. But they're seldom equipped with more capability than they absolutely need. Table 1 shows the power consumption for common WPAN technologies at their standard operating distances.^7,8,9 Note that while IEEE 802.11b is a wireless-LAN technology, it is included in this table for comparison with the WPAN technologies.

Another key constraint for mobile-device designers is bandwidth. Different usage models have differing bandwidth requirements. For example, real-time audio can be reasonably achieved with relatively low bandwidth. So for effective real-time audio, raw throughput may actually be counterproductive. Smooth responsiveness is the feature that's most needed for real-time audio. This responsiveness may be achieved with reasonably low transmission rates. In contrast, moving a photograph wirelessly from a megapixel digital camera requires a fairly generous bandwidth.

Generally speaking, handheld devices tend to provide relatively low bandwidth when compared to larger computing platforms. As an example, Bluetooth provides a capability of 1-Mbps raw data. This data rate can be maintained simultaneously by multiple piconets, depending on the number of collisions that occur during frequency hopping. But in an environment that's saturated with RF traffic in the 2.4-GHz band, each Bluetooth device may have available bandwidth that is considerably less than this amount.¹⁰

In contrast, IrDA infrared devices can achieve fairly high bandwidth (up to 16 Mbps) in a very small footprint with relatively low power consumption. Even slower serial-infrared (SIR) transceivers accommodate dedicated links between devices. They achieve speeds of up to 115.2 kbps using off-the-shelf serial UARTs.¹¹

In addition to power and bandwidth, cost remains one of the most critical constraints for mobile wireless devices. WPAN devices tend to be more cost constrained than other technologies. For instance, an IrDA infrared transceiver may cost around $1 per unit. Bluetooth transceivers are more expensive (currently around $20 with hopes of falling to $5 in the next few years).¹² Wireless-Ethernet transceivers are even more expensive. For example, the cost of an IEEE 802.11b-compliant unit ranges from $50 to $100.¹³

In today's market, secure systems are becoming mandatory. As a general rule, the less connected a device is, the more secure it will be. Obviously, security must be balanced against access. If data is in a sealed box and buried in the ground, it's pretty secure. But it's also of little value. For WPANs, however, limited range and specific usage models certainly provide greater security. If two devices are interacting via a 1-m infrared link, a security breach would have to occur within arm's reach of the individuals who are sharing information. Of course, a breach also could occur if the reflected light is detected and the surrounding ambient noise is filtered out.⁶

Due to the physical characteristics of the transmission medium, short-range RF solutions aren't as secure as infrared. But they still provide a relatively limited range and hence a limited opportunity to breach security. Due to the insecurity that's inherent in the Bluetooth physical layer, Bluetooth implements link-level security like authentication and encryption. Some critics feel that Bluetooth's security measures aren't sufficient. In particular, they feel that link and encryption keys can be stolen by exhaustively searching through PINs or mounting a middle-person attack.¹⁴

In summary, WPANs tend to be cheaper, smaller, simpler, more secure, and less power hungry than the technologies that provide infrastructure access. They also tend to have lower bandwidth and shorter range. The key driver in WPAN design is the suitability of a protocol to a particular purpose (or usage model). In addition, users benefit from the ease with which these local wireless devices can dynamically configure to meet their needs. With WPANs, it's essential to remember that the focus isn't a single technology, but rather a set of technologies. Each one is adapted to the particular usage models that are required for a variety of devices.

While the future will undoubtedly bring advances in WPAN technology, two solutions are currently available for consumer products: Bluetooth and IrDA. These two technologies provide a study in contrast. Bluetooth is an RF technology that operates in the 2.4-GHz ISM band. It provides the ability for up to eight devices to form a dynamic piconet. To avoid interference from other radio sources, Bluetooth employs rapid frequency hopping between 79 channels. Such interferers include microwave ovens, 802.11, HomeRF, and other Bluetooth devices. Currently, Bluetooth provides a range of 10 m at raw speeds of up to 1 Mbps. In the future, it promises to increase both that distance and the bandwidth. Presently, Bluetooth's greatest strength may be its grassroots support (with approximately 2500 SIG members). It also has a highly effective marketing machine.

As a WPAN technology, Bluetooth shines when users need mobility within a somewhat limited space (10-m sphere). It also flourishes when relatively low data bandwidth is acceptable for the usage model. Bluetooth is largely targeting consumer electronics. It includes usage models for mobile headsets, remote printing, cell-phone-to-laptop connectivity, and cell-phone-to-cell-phone connectivity.

In contrast, IrDA is a standard for data communications from the Infrared Data Association. Using pulses of infrared light, it communicates at speeds as high as 16 Mbps over distances of up to 1 m. IrDA defines a cone of light with a minimum 15° and maximum 30° half angle. Hence, it is a "point-and-shoot" technology.¹¹

IrDA is ideal for situations in which dynamic ad-hoc connections need to be made quickly. It also shines when high data rates are needed or privacy is a necessity. In all cases, the usage models that involve IrDA must be able to work well under the constraint of a very short distance. Effective usage models for IrDA include wireless point-of-sale (POS) terminals, point-and-shoot object push (including the exchange of objects between PDAs and laptops), and the movement of large data (such as the transmission of pictures from megapixel digital cameras).

Since the arrival of Bluetooth, an interesting question has been whether the future would bring a peaceful and harmonious cooperation between these two WPAN technologies. As some have posed, will it be a battle to the death?¹⁵ In my opinion, the initial rumors of IrDA's death were greatly exaggerated. The initial hype seems to be giving way to an understanding. While a definite overlap does exist between Bluetooth and IrDA in certain usage models, there are other usage models for which one clearly outperforms the other. Table 2 highlights the fundamental differences between Bluetooth and IrDA. It suggests potential synergistic solutions when complementary strengths are combined in a single device. The future should hold increasing cooperation between these two technologies. In some situations, both Bluetooth and IrDA transceivers will co-exist in the same device. They will leverage their respective strengths to create an even better user experience than is currently possible with either technology alone.¹⁶

In conclusion, wireless personal-area networks encompass a number of usage models that involve wireless data movement within a relatively short range. For these usage models, interoperability (rather than access to an infrastructure) may be the killer application. A wireless-headset user may care very little about whether his or her headset can access the Internet. But that user will care a great deal if that headset can't play music from the laptop and the stereo or roam between the two. The primary value of WPANs lies in mobile devices doing meaningful things together while operating independently of a network infrastructure. For those WPAN devices that need access to the network, access-point capabilities are an available and valuable usage model. When it makes sense to do so, WPAN devices may access an available network infrastructure. When such network connections aren't required, however, they simply operate independently.