Convenience is the major motivation behind our adoption of wireless technology. The ability to access information anytime from anywhere drastically increases our productivity as well as our quality of life by allowing us to work from home, car, school, or vacation resorts, and manage time more flexibly. Besides the unprecedented convenience, a wireless Internet also improves the quality of the information services. By taking advantage of wireless devices' pervasive nature, you can dynamically customize information services for each user based on her location, mood, or other real-time data.

The wireless Internet's dynamic nature requires a new breed of dynamic services, which likely involves many service providers. Even simple mobile commerce tasks require much expertise. For example, a straightforward stock trade application necessitates services such as user authentication and authorization, customer relationship management, stock quotes and tracking, trade executing, and fund transfer. Those services are built upon more basic services, such as Internet connectivity, security, and transactional reliability monitors. The IT industry's history has shown that a single vendor cannot offer the best expertise or services in all fields. The best economic model features small vendors providing modularized service components, each serving its core competency. Those modularized reusable software components are called Web services. To minimize bandwidth and CPU usage, wireless users can selectively use only the core services they need. Applications that utilize a wireless front end for pervasive human interfaces and Web services on the back end to leverage the Internet's vast information resources are called wireless Web services applications.

Web services from different vendors must interoperate from machine to machine to provide transparent services to customers. Distributed computing services have long been able to work with each other through remote procedure calls (RPC) and remote object frameworks. However, building such an interoperable network on an Internet scale requires industry-wide standardization of communication protocols. We have XML-based protocols to standardize Web services' dynamic discovery processes (UDDI, or Universal Description, Discovery, and Integration); service interfaces (WSDL, or Web Services Description Language); and asynchronous and synchronous messaging (SOAP, or Simple Object Access Protocol, and XML-RPC). On top of those communication protocols, other protocols provide more enhanced functionalities, such as the XML Encryption and XML Digital Signature protocols for improving security. For instance, you can embed an XML digital signature in a SOAP message to guarantee that message's integrity.

Throughout this article, we explain wireless Web services' network architectures, how Java fits into the wireless Web services picture, evolving technologies, and future trends. Since much literature already discusses server-side Java platforms and Java-based Web services, we focus only on Java's role in wireless application development.

Wireless Web services architectures

A wireless Web service can feature one of the following architectures: wireless portal network, wireless extended Internet, or wireless ad hoc network.

Wireless portal network

In a wireless portal network, the wireless information devices connect to the Internet backend services through portal entry points. The portal creates a "walled garden" and controls access to Internet contents. Wireless portal networks support widely deployed thin-client wireless technology, such as WAP (Wireless Application Protocol). In general, a WAP-based Web services portal works like this: The wireless device sends an information request embedded in a WML (Wireless Markup Language) form. The portal receives the message, checks the user's privilege, and then translates the request to a SOAP message or an XML-RPC call to an appropriate partner Web service. The Web service replies, and the portal translates the response back to a WML document. The portal sends the WML document back to the wireless device for display. In this way, the portal works as a proxy for wireless users. The portal operator provides user authorization and management services. Many partner vendors can provide real application Web services under the ASP (application service provider) model.

The advantages of thin client-based portal networks include:

  • The technology is mature, widely deployed, and well supported. If you require immediate wireless capabilities, you should investigate this option.
  • Thin clients require little computing power and therefore can run on small, light devices.
  • Only server-side computers can handle many computationally intensive application functionalities—such as voice recognition engines, as we will discuss later.

However, wireless portal network combined with thin clients might not fit best for dynamic wireless Web services for the following reasons:

  • From the business viewpoint, the portal architecture requires a central entity to control all user data. Backend service providers must rely on the portal architecture to access users. The portal provider decides what services can access its walled garden and demands a profit share from partners. That could cause serious business concerns if an untrustworthy, insecure, or abusive monopoly controlled the portal.
  • The portal architecture has a central failure point, which exposes the whole network to attacks.
  • Thin clients offer limited user interfaces. They are not pervasive.
  • Thin clients do not remember complex application state information; thus, the user must navigate through multiple pages to complete complex tasks, such as transactions. Not only is that inconvenient to the user, but it is also slow and uses precious wireless bandwidth. Moreover, it renders thin clients vulnerable to network interruptions.
  • Thin clients only support limited and predefined security protocols, such as HTTPS secure connections. We cannot have application-specific flexible security policies. In case of a WAP portal, a misconfigured gateway can easily cause security gaps.

New advances in wireless devices technology allow small devices to run more sophisticated client software, or fat clients. A network structure designed for fat clients can eliminate many problems of the thin-client portal network solution.

Wireless extended Internet

Wireless extended Internet is the wired Internet's expansion to wireless devices. Today's handheld devices have more computing power than many desktop PCs did 10 years ago. Wireless information devices can have their own IP addresses (through Internet Protocol 6) and full network functionalities. Those devices usually run smart, fat clients that interact with multiple backend services simultaneously and store/process application data on the device.

Smart devices support sophisticated user interfaces, offline processing, and automatic transactions. They can also implement flexible, application-specific security policies. Like the Internet itself, the wireless extended Internet architecture is decentralized (without a controlling business entity) and eliminates any single point of failure. However, as you will see later, centralized Web services hubs are still required to support advanced security schemes (such as single sign-on authentication) and user interfaces (such as VoiceXML engines). Unlike the portal architecture, the hubs themselves can be decentralized. Different vendors can provide similar hub services that can interoperate with each other. Figure 1 shows a topography diagram for such networks.

Figure 1. Wireless Web services architecture. All communications use XML-over-HTTP protocols. Server-side Web services group into single sign-on security realms. In reality, hubs other than single sign-on authenticators exist.

The extended wireless Internet architectures blended with decentralized hub Web services will provide the foundation for future wireless Web services applications, an approach we focus on throughout this article. Since most of the supporting technologies are just emerging, many challenges prevail. In later sections, we will discuss those challenges and suggest solutions.

Wireless ad hoc network

The wireless ad hoc networks allow wireless devices to become servers to peers. Wireless peers can provide content, network traffic routing, and many other services. The ad hoc network truly leverages wireless networks' dynamic nature. However, because wireless peer-to-peer technology is still embryonic, its many performance and security issues must be solved before it can be widely used.

We do not discuss wireless ad hoc networks in detail in this article. Readers interested in Java-based peer-to-peer frameworks should check out Jxta, Sun Microsystems' open source peer-to-peer framework. Jxta can run on both server computers and handheld devices.

What is Java 2 Platform, Micro Edition?

In addition to choosing the most appropriate wireless Web services architecture, we must also choose the most powerful and flexible development platform. In this section, we examine the Java development platform for wireless applications—Java 2 Platform, Micro Edition (J2ME).

Though a lightweight version of Java 2 Platform, Standard Edition (J2SE), J2ME is not exactly a subset of J2SE. J2ME keeps some of the J2SE core library APIs, but substitutes others with lightweight components through the javax.microedition package. Sun, understanding the large variety of wireless devices and realizing that one size does not fit all, divided the J2ME platform into several layers and groups.

At the bottom of the J2ME hierarchy are two configurations that provide JVMs and core language libraries. Connected Limited Device Configuration (CLDC) is for the smallest wireless devices with 160 KB or more memory and slow 16/32-bit processors. CLDC has limited math, string, and I/O (input/output) functionalities and lacks features such as JNI (Java Native Interface), custom class loaders, and reflection. As a result, CLDC virtual machines cannot support most J2SE standard libraries.

Connected Device Configuration (CDC) is for more capable wireless devices with at least 2 MB of memory and 32-bit processors. CDC supports a fully featured Java 2 VM and therefore can take advantage of most J2SE libraries.

Configurations provide the most basic and generic language functionalities. On top of each configuration are multiple profiles targeted at specific devices. The profiles provide more advanced APIs such as a graphical user interface (GUI), persistent storage, security, and network connectivity. Mobile Information Device Profile (MIDP) and PDA Profile, two profiles based on CLDC, target cell phones and PDA devices, respectively. Based in CDC, the Foundation Profile provides more utility, network, and security functions for embedded devices, but no GUIs. The Personal Profile sits on top of CDC and the Foundation Profile. It provides J2SE Abstract Windowing Toolkit (AWT)-compatible GUI APIs for high-end PDAs and wireless Internet appliances. Figure 2 illustrates the J2ME platform architecture.

Figure 2. J2ME components and architecture

Though a lighter version of Java, J2ME still retains the language's underlying benefits.

The Java advantage

We all know the benefits of the Java platform; let's see how they translate to wireless Web services development.

Advanced language features and standardization process

Java's modular nature allows it to expand and develop solutions for new computational problems. It has evolved from a popular client applet language, to a cross platform GUI builder, to an application server platform. This same modular nature now allows Java to drive wireless and Web services applications.

Java also offers a democratic standardization process. To attack specific problems, individual vendors often initially develop Java libraries, which can result in inconsistent APIs for similar functionalities. To keep Java applications interoperable and Java code portable in the future, we must consolidate and standardize APIs. The JCP (Java Community Process) ensures that all interested parties voice their concerns and that no standard API unfairly benefits any one vendor. Standardization proves especially important for wireless Web services, due to the large number of vendors and providers involved. Many organizations have already started developing major JSRs (Java Specification Requests) that will standardize important new tools and APIs for wireless and Web service Java applications. We will mention those ongoing efforts later in this article.

Cross-platform compatibility

Windows-dominated desktop computers might not leverage Java's Write Once, Run Anywhere compatibility, but Java's platform-independence could benefit both wireless frontend applications and Web services backend applications.

Multiple hardware and OS platforms have and will continue to share the wireless information devices market for the following reasons:

Page 2 of 3
  • Due to their personal and pervasive nature, wireless devices are perceived as consumer electronics. Consumers do not care what is under a device's hood as long it does the specific job, thus allowing multiple vendors to co-exist in the same product category. Each vendor can achieve similar functionalities using different approaches. For example, different vendors manufacture dozens of cell phones on different platforms that all provide largely the same functionalities.
  • Unlike desktop computers, which are designed for generic computing tasks, wireless devices usually complete specific tasks with minimum sizes and optimal user interfaces. Thus, the huge variety of mobile applications requires many different devices and innovative solutions from many different vendors. No single vendor can provide the best solution for all tasks. For example, among PDA vendors are a variety of CPU and OS platforms. Palm OS, the most popular PDA platform, targets small, cheap, easy-to-use devices with basic features. Windows CE and embedded Linux devices are quickly gaining market share on high-end devices.

The heavy competition in the wireless devices market benefits customers by offering choices. But the fragmented platforms create a nightmare for developers. Repeatedly developing and maintaining different versions of the same application for different devices is prohibitively expensive. Yet, any general-purpose wireless application must run on all types of devices customers might own. Thus, cross-platform compatibility is a core issue to consider when developing wireless applications.

The server back end features many platforms ranging from Linux, to Unix, to Windows. J2EE's (Java 2 Platform, Enterprise Edition) success has proved the value of cross-platform compatibility on the server side. As the back end evolves into multivendor, multiplatform, and interoperable Web services networks, Java's value will increase.

Though a core advantage of Java, cross-platform compatibility comes at a price: a performance price. No absolute compatibility exists among a huge variety of wireless devices. For example, you can't possibly port the full features of an application designed to run on a TV set-top box to a tiny cell phone. Even among similar devices, such as high- and low-end cell phones, portability can cause under-utilization of resources on one device, while straining the resources of another. The J2ME platform balances Java's portability with performance on many wireless platforms. As we discussed, Sun divided J2ME into multiple configurations and profiles to achieve maximum functionality and performance for different device categories, and ensure portability within any one category.

Page 3 of 3

Security

In addition to providing a compatibility layer, the JVM also provides application security through the following ways:

  • The JVM verifies the byte-code application at class load time to ensure that the application does not do anything dangerous
  • The JVM monitors the application execution and avoids memory leaks
  • The JVM can incorporate a security manager or sandbox to verify application digital signatures and grant permissions to access specific APIs (domains) based on the signer's trust level

The three security features above prove important to wireless applications. Wireless devices usually store personal and sensitive data. Users often download mobile code from an insecure Internet. We must ensure that the downloaded application is not a virus and will not try to corrupt or even steal data from the device. To combat such attacks, the J2SE and J2ME/CDC platforms offer fully featured, secure JVMs. J2ME/MIDP uses a simplified byte-code verification scheme to minimize performance hits. Domain-based security managers are available for MIDP 2.0 VMs.

In addition to application security, Java provides excellent support for industry-strength authentication, authorization, and cryptography-based communication security protocols. The Java Cryptography Extension (JCE) and Java Authentication and Authorization Service (JAAS) frameworks are both pluggable and flexible. They provide a set of abstract interfaces of common security algorithms and concepts to application developers. Different providers supply the implementations. You can easily add new algorithms into the frameworks. Java's security framework design separates applications from security solution providers. Therefore, it allows you to choose or switch to any provider and algorithm at any stage of software development without learning new APIs or worrying about compatibility issues. We will discuss more about Java application security technologies in the following section.

Core technologies

We now discuss the core Java technologies needed to support wireless Web services. Some are already available in standard Java APIs; some are currently offered by third-party vendors and are being standardized; others are only in the planning and designing stages.

Interoperable XML standards

As we mentioned earlier, Web services are built around standard XML communication protocols. In wireless Web services applications, servers and clients are loosely coupled with XML messages. So both Web services application servers and wireless devices must support standard XML protocols.

On the server side, the J2SE/J2EE platform offers excellent support for generic and Web services-specific XML processing through the Java API for XML Parsing (JAXP). JAXP features the following:

  • Both pull and push mode SAX (Simple API for XML) parsing
  • Standard DOM (Document Object Model) and a more Java-friendly JDOM format
  • Java API for XML Binding (JAXB), which supports persistent XML mapping for Java objects
  • Java API for XML Messaging (JAXM) and JAX-RPC, which can support Web services SOAP and XML-RPC messages
  • Java API for XML Registries (JAXR), which supports XML registries such as UDDI and ebXML

On the wireless front, mainly three concerns crop up when using XML on small devices:

  1. CLDC VMs have limited string functionalities and cannot run generic XML parsers written for J2SE or any Web services XML APIs built on top of them
  2. Support for XML requires increased processing power
  3. XML formats embed data in verbose text tags, which does not conserve limited wireless bandwidth

Third-party groups offer many CLDC-based XML parsers that require little processing power. Most, such as NanoXML and TinyXML, only support the simpler serial SAX-style parsing. However, in-memory objects are required if you want to manipulate complex XML data. Package kXML developed by Enhydra.org offers both SAX- and DOM-style parsers. A W3C (World Wide Web Consortium)-compliant DOM parser proves too heavy for small devices, so kXML uses a lightweight alternative called kDOM. In addition to generic XML parsers, kXML also has packages specialized for SOAP (kSOAP) and XML-RPC (kXML-RPC) message processing. With JSR 172, the J2ME Web Services Specification, the JCP continues its effort to standardize XML support on the J2ME platform.

XML's excessive bandwidth usage has long proved problematic in wireless application development. WAP developers use an XML compression protocol called WBXML (WAP binary XML) to compress WML communication data. You can adopt WBXML as a standard format for compressing generic XML documents. You must install matching compressor and expander modules on both wireless and server sides. kXML provides both modules.

Wireless user interfaces

The major advantage of wireless information devices is their pervasiveness; they melt into our everyday lives. This ability requires not only small, wearable devices, but also devices with nonintrusive user interfaces. The success of Palm PDAs has proven the effectiveness of touch-screen and handwriting recognition input methods. Migrating mature desktop GUIs to small devices has also proven successful since today's computer literate public needs little training to use common GUI components such as buttons, selection boxes, and textboxes. Most J2ME technologies support GUIs. The smallest MIDP has its own lightweight GUI components and event-handling models. More complex profiles, such as the PDA profile and Personal Profile, adopt GUI models commonly used in the desktop J2SE world (such as AWT) to ensure the maximum portability level.

Though convenient, GUI-based devices are still interruptive. For example, on a GUI-based device, users have to use both hands to input long text or make complex selections. By contrast, voice-navigated information services can integrate into workflows much better. We can talk on the phone while performing other sophisticated tasks (such as driving). The leading technology enabling voice-based information navigation is VoiceXML. Similar to HTML, a VoiceXML page can present information, take user input, and jump to another page based on that input. The difference: An HTML page presents and takes text information, while a VoiceXML page handles voice information. However, VoiceXML pages are written in XML text format since backend applications that generate VoiceXML pages only work with text data. Voice recognition and synthesis technology proves crucial to bridging voice and XML text formats.

JSR 113, the Java Speech API 2.0, proposes a set of standard Java APIs for voice recognition and synthesis. But voice algorithms still prove too computationally intensive for small handheld devices and require expensive software. Current VoiceXML applications require gateway servers to perform voice recognition and synthesis. Those voice gateway servers could become parts of voice portals, similar to WAP portals. The voice portals act as information call centers for customers: they contact Web services on behalf of users and translate information back and forth among voice, text, VoiceXML, and SOAP messages. However, with the exclusive voice portal model, the mobile application can only access the portal through a voice phone line/channel, and all information must be presented in voice format.

For many applications, we want to mix voice information with other forms of digital information, such as graphics, text messaging, and streaming video. One way to integrate voice into TCP/IP network-based applications is through Voice over IP (VoIP). VoIP specifies how we should transport voice on a data network. VoIP-enabled VoiceXML gateways can respond to voice streams from the Internet instead of a phone line. A VoIP server can be just a hub service in the vast network of backend Web services.

Figure 3 illustrates the architecture of a smart wireless application that uses both voice and common GUIs.

Figure 3. Mix voice, text, and graphics in one wireless application

Security

As discussed previously, Java supports cryptography and mobile code security. However, securing Web services proves more challenging than securing traditional Internet applications. Web services are designed to improve productivity and interoperability; security comes as an afterthought and is largely vendor-specific. Yet, Web services run outside corporate firewalls and can potentially expose an internal network to the outside world. As a result, security concerns provide major obstacles in deploying Web services. Security is also a chief challenge for wireless applications due to radio signals' vulnerability to interception and the lack of computing power for performing strong cryptography operations on small devices. In this section, we discuss Java technologies on both server and wireless ends that improve wireless Web services security.

Web services require end-to-end security for XML messages traveling across multiple intermediaries, thereby making traditional point-to-point secure communication channels (such as HTTPS) inadequate. Secure XML protocols such as XML Digital Signature and XML Encryption can bind with Web services communication protocols (such as SOAP) to secure XML contents. JSRs 105 and 106—the XML Digital Signature APIs and XML Digital Encryption APIs, respectively—propose Java APIs to support XML Digital Signature and XML Encryption. However, on the wireless front, most standard Java security and cryptography APIs are only available to J2SE and J2ME/CDC applications. Due to small devices' limited computing power, J2ME/MIDP only officially support HTTPS secure connections from the upcoming MIDP specification version 2.0. Only third-party vendors provide support for general cryptography algorithms on the J2ME/MIDP platform. One such lightweight CLDC-compatible cryptography package is the Bouncy Castle Crypto APIs. To learn how to use these APIs, please refer to Michael Yuan's developerWorks article, "Securing Your J2ME/MIDP Apps," (June 2002).

Another vitally important security feature in Web services is single sign-on, required for keeping multiple service vendors transparent to users. Sun's Liberty Alliance Project promotes single sign-on services based on federated identity. The alliance will probably base the implementation on Kerberos technology. You can deploy Kerberos plug-ins for JAAS for tight integration among Web services. You can also embed Kerberos-like security tokens in SOAP message headers as SAML (Security Assertion Markup Language) elements for loosely coupled solutions. JSR 155, the Web Services Security Assertions, proposes Java APIs that support SAML. IBM and Microsoft's WS-Security, another set of XML protocols, competes with SAML. The IBM Web Services Toolkit offers Java solutions for WS-Security.

When approved, JSR 177, the Security and Trust Services API for J2ME, will provide Java APIs to SIM (Subscriber Identity Module) cards embedded in wireless devices. SIM cards usually store identity and cryptography keys. They provide automatic authentication and authorization on wireless phone networks. Together with SAML or WS-Security, JSR 177 APIs could enable automatic single sign-on services for SIM-enabled wireless devices.

For more discussion on J2ME and Web services security, please refer to our article "Securing Wireless J2ME" (developerWorks, June 2002).

Performance

One of the biggest problems with smart wireless applications is performance. As alluded to earlier, Java's development productivity and cross-platform compatibility take priority, and, thus, the JVM cannot benefit from many hardware-specific optimizations. However, advanced security and application data-processing features require high performance on the wireless end. Efforts such as JSR 177 are underway to make special hardware features available to standard Java APIs. Technologies such as Java-optimized CPUs and coprocessors, and just-in-time Java compilers are already available. For many applications, Java programs run as fast as C/C++ applications. Sun's project Monty promises to deliver CLDC and CDC VMs that run 10 times faster than current VMs with small footprints.

Go wireless

In this article, we introduced wireless Web services and the Java technologies that could power them. Current and emerging Java platforms are well positioned to help develop pervasive wireless solutions. Today's buzz phrase, wireless Web services, is becoming tomorrow's exciting reality!

In subsequent column articles, we will discuss specific techniques to implement wireless Web services applications.

Michael Yuan and Ju Long are PhD candidates at the University of Texas at Austin. They are both research associates in the Center for Research in Electronic Commerce. Their main interests are mobile commerce and wireless-based marketing research.

Learn more about this topic