INTERNET APPLIANCES
Do Appliances Threaten Internet Innovation?
Sharon Eisner Gillett, William H. Lehr, John T. Wroclawski, and David D. Clark MIT Internet & Telecoms Convergence Consortium
ABSTRACT sumer still buys her own PC, but other portions of the overall package are bundled. Cable modem The world is evolving from one in which services, for example, bundle the cable-based net- almost all access to the Internet comes through work access with ISP services such as assignment personal computers, to one in which so-called of e-mail and IP addresses. The user gives up the Internet appliances are expected to make up a freedom to choose her ISP in exchange for the growing share of end user equipment. Focusing enhanced connectivity service (higher speed, on consumer-oriented appliances, we consider always on) broadband provides. AOL bundles whether this shift has implications for the pace even more, packaging network access together of Internet innovation. We conclude that given with AOL-specific application software and ser- the starting point of the current Internet, certain vices, such as an AOL e-mail account. In of the proposed business models for Internet exchange for the convenience of this bundle, the appliances are not likely to be viable, and this user gives up the flexibility to mix and match very fact protects the Internet’s characteristically components; for example, she can’t use non- rapid pace of innovation. AOL e-mail client software (e.g., Eudora) to read her AOL e-mail. NTRODUCTION At the other extreme lies a fully integrated I model exemplified by the various “free PC” ser- Today, and especially in the United States, vices that have been tried. In this model, the PC most users access the Internet using a personal is contractually bundled with ISP and application computer (PC). This is expected to change as services. In exchange for giving up her choice of ever more consumer-oriented, appliance-like which PC to buy and which service or service ele- devices start communicating via the Internet. ments (e.g., application software) to choose, the As a general-purpose computing device, the user gets a complete package that works out of PC has complemented the flexibility of the Inter- the box and can be set up without any error- net’s architecture to support a spectrum of busi- prone configuration or time-consuming research ness models for Internet services with varying among competing alternatives at many levels. degrees of vertical integration. Broadly speaking, Clearly, these different models appeal to dif- the more vertically integrated the service, the ferent types of users. Some people appreciate 1 In other words, the user easier it may be to use, especially for neophytes, how easy AOL’s e-mail client is to use and does not have to replace but the less flexibility it provides to users. To don’t want to be bothered with competing alter- her access device. Service date, multiple models have coexisted, serving natives. Others are accustomed to more power- providers, however, have groups of consumers differentiated by their rela- ful client programs from their workplaces and little incentive to make tive weighting of these attributes (Fig. 1). find it greatly limiting not to be able to access other aspects of switching At one extreme lies a fully disaggregated their personal e-mail in the same way. The key easy. For example, most model selected by consumers who place a premi- point is that the user has the choice of which ISPs, including AOL, will um on flexibility and freedom of choice. In this model to adopt depending on her individual not forward e-mail sent to model the consumer buys her own PC and phone preferences. Furthermore, because in all of [email protected] if line, selects an Internet service provider (ISP) these models the user access device is the same you leave randomISP.net from among many competitors, loads the soft- general purpose PC, the user can shift among behind. Surmounting this ware of her choice (any Web browser, e-mail the models reasonably fluidly.1 A free PC user barrier is primarily a mat- program, etc. conforming to Internet standards) whose comfort level rises as he uses the box has ter of effort (tell all your and configures it to visit whatever pages she the technical capability to take off the training correspondents your new wants, download e-mail from whatever server wheels and switch to more disaggregated ser- address, sign up for a free she wants, and so on. vice providers — as long as he is willing to wait email-for-life forwarding In the middle lie semi-integrated models such for the service contract to expire, or simply service, etc.), however, not as broadband Internet access delivered via cable treat the financial commitment to the free PC cost. modem or America OnLine (AOL). The con- provider as a sunk cost.
46 0163-6804/01/$10.00 © 2001 IEEE IEEE Communications Magazine • October 2001 Although not all Internet users will select the fully disaggregated model, the fact that some do helps drive the innovation engine of the Internet. Dis-aggregatedStructure of service Bundled For example, several years ago non-AOL e-mail clients introduced the capability to attach Web pages to messages and view them from within the e-mail program. As a result, it became common practice for people who used such clients to send Earthlink AOL FreePC Web pages around via e-mail. AOL users who received such messages began to want this capa- bility as well, eventually leading AOL to integrate this feature into its e-mail client. FlexibilityWhat customers value Ease-of-use This example illustrates the critical lead user role played by subscribers who favor the disag- gregated service model (those on the left side of Fig. 1). Because these subscribers are the most Figure 1. The spectrum of business models supported by the PC as an Inter- comfortable with installing new software on their net access device. PCs, they are the most willing to experiment with new Internet services and applications such as Internet telephony or peer-to-peer file sharing. appliance, according to who chooses the particu- The suppliers of such innovations, such as Vocal- lar set of tasks: the manufacturer (Class 1), a Tec and Napster, count on such lead users to try provider of services required to use the appli- out new ideas and to help motivate their diffu- ance (Class 2), or the user (Class 3). sion to the larger mass of users who prefer a Class 1 devices are the closest to traditional more controlled experience. Lead users, in other appliances like toasters and refrigerators. They words, are an essential part of the demand that can only do exactly what the manufacturer built spurs the ongoing supply of innovation in the them to do. Portable scanners that can upload networked environment of the Internet. images over the Internet, cell phones enhanced 2 See [1] for more on cre- If it is true that the faster the pace of Inter- with built-in Web browsers, and devices that play ative destruction net innovation, the more rapid the process of Internet radio stations are examples of Internet unleashed by the Internet. creative destruction that drives productivity improvement and economic growth, then devel- opments that threaten to retard Internet innova- Primary Example products/concepts tion should be matters of concern for public characteristic policy.2 The central policy issue explored in this article is whether Internet appliances constitute Portability • Portable scanner (HP Capshare) such a development, or whether they simply rep- • Walkman-style MP3 player resent a beneficial expansion of Internet cus- • Voice-data convergence: tomer premises equipment (CPE) options. To – Cellular Æ Internet (Europe: WAP; US: Sprint address this policy question, we must first define Wireless Web) what we mean by an Internet appliance. – PDA Æ Internet (Palm VII; Palm III or V with Omnisky (CDPD) modem; (Visor with phone plugin; WHAT IS AN INTERNET APPLIANCE? KyoceraSmartphone) As Table 1 shows, widely varying product con- – Pagers Æ email (Motorola Talkabout; RIM Blackberry) cepts have been lumped under the “Internet • General wireless data (Japan: NTT Docomo’s iMode; appliance” rubric, including entries as diverse as Europe/US: 2.5G, 3G, 4G) mobile e-mail devices for power users, ther- Multimedia • Digital picture frames (Ceiva; Storybox) mostats controlled from Web browsers, and non- entertainment • Personal video recorders (TiVo; Sonicblue, formerly PC computers designed to run network-based ReplayTV) applications. Although advances in portability, • Net-top boxes (WebTV, AOLTV) ease of use, cost, and so on are important gener- • MP3 jukeboxes ators of new devices, each of these characteris- • Internet radios (3Com Kerbango; AudioRamp iRAD-T) tics is not found in every Internet appliance. Rather, the common attributes that characterize Simplicity • Internet access without the PC (Netpliance i-Opener, Internet appliances are: Universal Talkware; Gateway/AOL Internet appliance; • Connection to the Internet, whether for a Compaq MSN Companion) few minutes each day or all the time, • Screen phones (Infogear/Cisco iPhone; Alcatel Webtouch) whether as first-class Internet devices or via • Email-only boxes (Cidco Mailstation; Landel Telecom translation gateways Mailbug) • Particularity of purpose Embedded • Internet refrigerator (Electrolux Screenfridge) Particularity of purpose means that some- Internet • Home security/environmental systems controlled via where before the end user’s experience of the Web-based user interface (Honeywell WebPAD) device, its function is fixed. Particularity of pur- • VCR with no knobs, controlled via Web browser pose is not typical of PCs. Users decide what to do with their PCs by choosing what software to Beyond Wintel • Thin clients install on them — including the possibility of • Wireless Web tablets (Quibit; Mitsubishi; Honeywell software developed after the PC was manufac- WebPAD) tured. An Internet appliance, in contrast, is designed to perform a particular set of tasks. We define a taxonomy of three classes of Internet Table 1. Diversity in the space of Internet appliances.
IEEE Communications Magazine • October 2001 47 appliances that could in theory be built as Class DO APPLIANCES Rather than 1 devices (although most aren’t in practice, for THREATEN INTERNET INNOVATION? reasons we discuss below). requiring users to A Class 1 Internet appliance has no capacity What effect are Internet appliances likely to take explicit to change its function once it has been built. In have on the pace of Internet innovation? One contrast to the PC, it has no hardware or soft- way to approach this question is to assume that actions to keep ware customizability: no slots for cards that the cost of innovation is a reasonable proxy for could modify or add to its functionality, and no its pace: ceteris paribus, the more innovation their devices capacity to accept new software. It may have costs, the less it will happen, and the more slow- current with system-level customizability if the manufacturer ly change will occur. Then we can ask how Inter- equips it with plugs to connect to other devices, net appliances of each class are likely to affect every possible but the nature of these plugs cannot be the cost of innovations large or small — whether upgrade, a Class changed. the introduction of entirely new applications or Class 2 Internet appliances, in contrast, can usage modes for the Internet (e.g., peer-to-peer 3 appliance might be changed after they are built, typically by soft- file sharing), or simply new features for existing ware download. Unlike with PCs, however, the applications (e.g., the ability to view a Web page be equipped with choice of which software to download and when within an e-mail client). software that is to install it is not made by the user. A distinct In a networked system like the Internet, inno- feature of Class 2 devices is their business model, vation incurs three types of costs: development, intelligent which includes not only a device but also a ser- distribution, and adoption. Development involves enough to vice from a particular provider. The service may the cost to conceive a new idea and implement it be basic connectivity (such as ISPs provide) but once (e.g., the cost to develop new software for observe and more often includes content-oriented features the PC). Adoption involves the cost to end users (such as America OnLine provides) as well. of picking up the idea (e.g., by modifying or interpret user Examples of Class 2 Internet appliances replacing their CPE). Distribution involves the behavior, abound: they include the Netpliance i-Opener cost of changing network infrastructure, if need- (originally designed to work only with Netpli- ed, to support the new idea. Distribution determine when ance’s ISP-like service); AOL and MSN compan- includes all costs (other than development) that an upgrade or ion devices built by Gateway and Compaq, must be incurred in parts of the networked sys- respectively; e-mail-only devices such as Landel tem not controlled by the end user, before the enhancement is Telecom’s Mailbug and Cidco’s Mailstation; dig- first user can try out the innovation. ital video recorders such as TiVo’s and Son- We expect development costs to remain needed, and icBlue’s (formerly ReplayTV); and digital picture unchanged by the introduction of appliances: the make it happen. frames such as the Ceiva and Storybox. Each of cost to develop an innovation will remain the cost these devices shares the characteristic that it to implement it in PC software. This outcome is a only continues to work as long as the owner con- result of the starting point that history has given tinues to subscribe to the associated service from us. The Internet already exists with most of its the particular service provider. This service users accessing it through PCs. In particular, the provider in turn determines the functionality of lead users who are the primary audience for the box, by controlling which software runs on it Internet innovation are PC users and likely to and when, if ever, that software is updated. remain so. Although they are likely to be the Finally, we posit a third class of appliances early adopters of Internet appliances as well, early that do not currently exist. Class 3 appliances appliance successes (e.g., the Palm Pilot and would be like Class 2 in that devices could be Blackberry) and failures (e.g., Netpliance) suggest updated, but unlike Class 2 and like PCs in that that these users’ focus will be on devices that aim users would control the updates. Class 3 appli- to complement the PC rather than substitute for ances would differ from PCs, however, in requir- it. Therefore, we do not expect the introduction ing less sophistication and involvement from the of appliances to change the lead user model for user to accomplish updates. They may also sup- the diffusion of innovation that we discussed port a more limited range of functions than a above, in which the users of PCs who are most PC, much as a Palm Pilot currently functions as comfortable installing new software remain the a sort of scaled-down PC.3 initial market for innovation. Rather than requiring users to take explicit Table 2 illustrates a framework for reasoning actions to keep their devices current with every about the effect that changing degrees of distri- possible upgrade, a Class 3 appliance might be bution and adoption costs have on the pace of equipped with software that is intelligent enough innovation in networked systems. to observe and interpret user behavior, deter- Today’s PC-based Internet fits in box 4 of 3 Although in principle a mine when an upgrade or enhancement is need- Table 2: it is characterized by low adoption and Class 3 device could have ed, and make it happen. PCs are already taking distribution costs, and therefore exhibits at least as much function- primitive steps in this direction, such as automat- extremely rapid innovation. Adoption costs are ality as a Class 1 or 2 ed installation of operating system updates and low because with the PC, users need only install device, it is possible that dialog boxes that automatically appear, asking software to adopt innovations. Such updates may certain elements that may users if they would like to download the plug-in carry some cost in complexity, but are typically be critical to the deploy- software they need to view content in a newly only incremental changes within a familiar ment of specific classes of released Web format. The development of meth- framework, with low monetary cost. Distribution applications may appear ods to more broadly, reliably, and proactively costs are low because innovations can be intro- first in Class 1 or 2 simplify the user experience through this kind of duced to Internet users with little or no change devices (e.g., GPS for implicit reconfiguration of the appliance is an to network infrastructure. location-dependent ser- active area of academic and industry computer In contrast to vertically oriented networks vices). science research. that were designed to support a particular appli-
48 IEEE Communications Magazine • October 2001 Distribution cost (cost imposed on network infrastructure to support first user’s use of the innovation)
High (vertically oriented networks) Low (Internet)
Adoption cost High 1. Slow (e.g., TV) 3. Rapid: non-viable appliance business (cost imposed on each user to model adopt innovation in CPE) Low 2. Fundamental innovations slow to emerge; 4. Rapid potentially rapid innovation at the margins (e.g., cell phone)