8.3 Appliance Design

8.3 Appliance Design

8.3.1 Form Follows Me

When designing smart devices, it is important to take some design issues into consideration before starting to build the device and the services around it. First, it is important to define the target groups and the related services. Based on this information, one should decide on a type of device that is adequate for this group (which could be, for example, a PDA, an Internet stereo, a refrigerator, or a Web pad). Next , it is important to understand what the related services for the product will be and in which environment it will be operated. Is it used, for example, indoors, outdoors, in a hot or cold environment? Depending on these requirements, the device will look different. Also of great importance is knowing of which context the application should be aware. Should it know its position, the outside temperature, or in which process step the user is at the moment? This will help to determine which additional sensors are required to make it easy to use. Appropriate context will make possible a great simplification of the user interface. Last, but not least, the target cost of the product should be clarified to make it commercially successful.

Currently, no one knows exactly what the customer will find useful. Manufacturers need to have adaptive designs and need to keep in touch with how the user interacts . It is important that a user does not have to learn a whole new way of interacting with the system, or it will fail (e.g., many older people who would benefit from buying a home PC are put off by an aversion to operating it). It is not enough to provide a mass of different devices for different user tasks , each connected to the network. The device must be easy, compelling, and enjoyable to use. Each device will compete others for space, mind share, and power outlets. As a consequence, some devices will be multifunctional, absorbing and displacing others.

Appliances must be designed from the perspective of the overall user experience. This includes the whole life cycle of ownership from initial advertising to purchase, installation, and use, to support, upgrade and eventual replacement.

Ergonomics play an important role in designing intelligent appliances. Only with good ergonomics will people use them. What needs to be taken into account is the human anthropometry in relation to workspace design. It is important to see how well the human and the appliance interact and how it can be improved for maximum performance. The arrangement of displays and controls needs to be verified and checked if they are to be linked properly.

Also to be taken into account are the cognitive and sensory limits of humans , which can make design unusable if the limits are ignored. Related to this, it is also important to check for fatigue and health issues that may arise from using the appliance. This is important when you design not only for the disabled, but for all users.

When designing the hardware, interfaces such as CRT and other display technologies need to checked for sensory and perceptual effects. You especially need to check legibility and display design. The interfaces need to be checked against the possible environment. These factors are even more true for speech input and output. If the device is meant to work outside, for example, you should make sure that you consider all possible weather conditions, including severe temperatures , winds, and storm conditions. You should also provide clear guidelines for its use under various environmental conditions, especially those that may be considered unsafe. If you use speech input and/or sound output, you should make sure that it can function under normal environmental noise factors and accommodate accordingly . Obviously, if you are designing a device to be used at a construction site, for example, you'll want to consider the optimal method for communicating with the device based on factors typical to such an environment. Likewise, additional care needs to be taken if you expect the devices to work in stressful or hazardous environmentsmake sure its design and endurance is appropriate to the situation in which it is expected to work.

Related to the interfaces, you need to make sure that the control design of the system is developed in a way that the users can use it without having to learn a lot. It means that the sequence of controlling the device is clearly structured and that you are only allowed to execute functions that are appropriate for this stage. The most favored devices will be ones that quickly acquaint themselves with their new masters, probably by borrowing knowledge and learned preferences from other appliances that have previously served the master well.

8.3.2 Technology Follows Me

Me-centric computing devices will be extremely diverse, including PDAs, set-top boxes, screen phones, in-vehicle computing platforms, smart phones, and home gateways, to name a few. As the market matures, this list of devices will continue to expand into areas that we cannot even imagine today. Described in this section are some of the requirements that must be addressed in order to build economically viable and successful devices.

Me-centric devices must appear to be appliances, not computers. This has implications such as instant on, user interfaces that match the function of the appliance, deterministic response times (even when disconnected), and clear and tight association between buttons, dials, and functions. A network-enabled set-top box's first priority is to deliver high-quality audio-visual content. The new functions associated with buttons and dials need to be as responsive as today's remote control when changing channels on a set-top box.

These devices target a broad range of price points, form factors, and hardware capabilities. Requirements vary across user interface models (including a range for both audible and small, medium, and large manual/visual), power requirements (battery life versus AC power), processor families (ARM, x86, PPC, MIPS, Hitachi SH, etc.), and network capabilities (fully connected, sporadically connected, and/or disconnected operations).

Human-computer interaction appropriate to the task and location needs to be sorted out. Liquid crystal display (LCD) touch screens, voice recognition, and text-to-speech output look like far better alternatives to the traditional PC with keyboard and mouse as methods of input and output. A consistent software user interaction model will also help achieve the wider goal of simplified learning and use. Also, it will encourage application development, thus avoiding software companies having to transfer their products to several different platforms. Existing desktop operating systems will not adapt well, as they depend on the present user input/output methods.

To be successful, it is necessary to rethink product concepts. While many manufacturers still think that the products they sell will create the profit the company wants to make, it is necessary to think of products as potential service offering platforms. These service offerings drive ongoing interaction with customers, which can become the cash-cow for the company. Eventually, it will be possible to give away the device for free, as the service fees will make sure that the company receives its share in the deal.

This has worked quite successfully with mobile phones. In Germany, the United States and Italy, for example, many mobile phones are sold for less than 1 euro/dollar or given away for free if the buyer subscribes to the phone network for at least two years . While the sale of the mobile phone would give the company a higher instant reward, selling subscriptions guarantees the company an income for the next two years.

To make this work, it is necessary to rethink existing business models. Ongoing annuity revenue becomes part of the picture, and customer relationships become a valuable asset in creating new partnerships, which build on brand recognition and trust. This creates opportunities to both offer services and to serve as a service integrator.

Through this new approach, it is possible to leverage increased customer interaction and data collection capabilities, which can lead to an increase in customer satisfaction. It also allows companies to provide improved products to the community and increases the company's partnership value and power lines.

When designing an appliance, it is important to make sure that one has the service in place before starting to design the hardware. Only this will guarantee that you reach the business goal that you are planning for (see also Table 8.2).

Table 8.2. Creating Value in Product Design

Important for me-centric appliances is the so-called "last mile" network connectivity to support the delivery of both data and application services. These include security models, device management, user management, content formatting, and software distribution requirements. Short-range ad hoc networking technologies, such as Bluetooth, HomePNA ^[3] , or WLAN based on IEEE 802.11 are becoming more common. This connectivity capability has implications on the device architecture requiring seamless support for transparent, appliance-like, ad hoc collaboration among these devices themselves.

^[3] http://www.homepna.org/

Table 8.3. Classes of Intelligent Appliances

Intelligent appliances can be grouped into four classes that address the complete landscape of requirements.

Thin Clients Render network delivered content to the user, either visually, audibly, or both with no local state information. Thick Clients Support applications using a programming model that is native to the device. Gateways Interface to the "last mile" of broadband capability while simultaneously providing the ability to host network-delivered services to be used on behalf of other devices in the network. Server Configurations Support the combination of gateway, thin terminal and thick terminal capabilities to allow, in a single device, hosted services, device-specific applications and content rendering for network delivered data.

{% if main.adsdop %}{% include 'adsenceinline.tpl' %}{% endif %}

Typically, intelligent appliances are grouped into four classes (see Table 8.3). These four classes address the whole range of requirements that we can foresee for the moment. They include thin clients, thick clients, gateways, and server configurations. The thin terminals do not contain any business logic themselves. They are used to deliver content, either visually, audibly, or both with no local state information. They act as a relay station for information and services. The advantage is that the footprint is rather small, but unfortunately these devices rely heavily on network connectivity, as they do not work without it.

Thick clients are like thin clients, but they have a larger memory footprint allowing the device to run applications and store data locally. This means that these devices can operate if there is no Internet connection. These devices support applications using a programming model that is native to the device. In many cases, Java is used. Gateway systems act as a local cache for thin and thick clients. They provide the gateway to the "last mile" of broadband capability. At the same time, these systems provide the ability to host network-delivered services to be used on behalf of other devices in the network. A server configuration supports the combination of gateway, thin clients and thick client capabilities to allow, in a single device, hosted services, device specific applications, and content rendering for network-delivered data.

To make me-centric solutions cheap enough for everyone to use, it is important to create reusable device architectures. This includes hardware components and a set of programming models that are used across device and network types to allow for the construction and deployment of device classes across the complete range of devices.

To connect an intelligent appliance to the network (Internet and LAN), discovery protocols need to be implemented that allow the appliance to search for networks and services that surround it. Once the appliance has found a network, it can connect to other appliances through LAN technologies, such as Bluetooth, HomePNA, or IEEE 802.11. At the same time, it will be connected either directly or via another local device to the Internet. While connectivity is desired, privacy can be an issue due to the lack of security. Not every device and service should be able to see each other, so smart cards, firewalls and encryption technologies are required for the authentication and authorization of users, agents , and services.

A more technical discussion will lead to questions on how much power consumption is necessary to keep operations alive for a time that is acceptable by the user. This will lead to the question of whether you can use a rechargeable battery or whether you need to plug the device into a power socket. This will define the range of use for a certain device, which can lead to major design changes.

The application will define the hardware requirements. If complex calculations are required, a faster CPU will be required that in turn consumes more power. The same question arises about memory and networking usage. Does the device require 4 Kilobytes or 512 Megabytes of memory? Does it need a slow infrared connection or a high-speed wireless LAN connection? This will have a huge impact on the power consumption and the design of the whole device. The same is true for output and input interfaces, such as displays and microphones and for sensors that help to give context to the device, such as GPS. Many other technical questions need to be answered before the device can be built, but in all cases the device should not be built because of the technical capabilities, but because of the requirements of the users and the proposed services.

8.3.3 Software Follows Me

Me-centric devices need content to deliver value. New devices cannot wait for entirely new custom content to be created in order to be successful. A cogent content architecture must address existing devices and lower the entry barrier for future devices seamlessly for all content providers in the value chain. This end-to-end content architecture also needs to ensure that new devices can be added while still supporting the network scalability demanded by millions of connected devices. Furthermore, as devices collaborate among themselves, the content needs to be naturally sharable and reusable.

Therefore the solution needs to use open standards that are not controlled by one entity and are easy to extend. It does not mean that everything needs to be open source, but the basic communication and application standards need to ensure compatibility of interfaces.

The software architecture needs to support both synchronous user request-response and asynchronous user messaging and notification. Depending on the application you are using and the requirements the users have, they will need both. The architecture needs to provide end-to-end scalability, which allows for growth in applications and user base. Applications can become more complex over time and may require more processing power and network resources. Besides end-to-end scalability, management is a very complex and important topic that needs to be taken into account. The management system supporting the solution needs to take care of users, devices, network-delivered applications, and content. Only through the management system is it possible to detect problems in this networked system. Just imagine a Web site that does not work anymore. The reasons for it can be manifold . It could be the device that you use to display it; it could be the browser, the connection, the server, the application, the database, the firewall, or the filesystem, just to mention a few reasons why a certain Web site cannot be displayed. Finding the problem that caused the downtime can be difficult if you have to do it manually. Therefore, automated management systems are required to manage the solution.

As the solution uses open standards, it will be possible to integrate third-party components, products, and adapters through well-defined interface points. This makes it possible to create more complex solutions that incorporate a range of different products and technologies from multiple software providers. Based on this software infrastructure, it is possible to reuse end-to-end network programming models. The supporting application model is based on components. These components work together in a cohesive manner to enable the construction of a wide variety of network-connected devices to address the device capabilities outlined earlier.

When designing the device, it is important to decide which operating system will run on it. While most applications will be written in Java, the different operating systems provide additional built-in features that are not to be neglected. Such features include speech recognition, applications, tools, control panel applets, custom shell, and handwriting recognition. This feature is bound to the device you are using and should be tightly integrated. Therefore, the decision for the operating system should be based on the additional services it provides.

Another important factor that needs to be considered in design is how new application code and content is downloaded to the device. Until now, most applications are transmitted via memory card or cable to these devices, which is slow and not very flexible. As more and more public WLAN access points are set up and GSM/UMTS-capable devices appear, it becomes clear that the content is transmitted on the fly wirelessly . But in some situations, it could be useful to put information on a memory card or to exchange information via memory card, in case there is no network connectivity available. Therefore, the device should support these memory cards from a software and hardware point of view.

From here on, you can start a detailed technical discussion about the software required on the device to make it run. Besides the operating system, you need a boot loader (a mechanism to load a program), a BIOS (the initial program that is loaded), and file systems, which would lead too much into technicalities. While you do not need to know the details, you should know that there is more you need to take into account, as these software bits and pieces require RAM, CPU power, and time to execute, which all can impact the design of the device.