Email	Print Reprint
	add to:
	Del.icio.us Digg Google Furl	Slashdot Y! MyWeb Blink

According to Forrester Research, more than a million Americans already own digital video recorders (DVRs). That number is expected to grow to 50 million by 2006, accounting for some 40 percent of U.S. homes. Though the utility of these devices has recently been overshadowed by national debate over content piracy, the fact remains that DVRs have the potential to completely change the way consumers control and manage their digital content.

To better exploit this potential, in January 2001, SONICblue's ReplayTV division launched an online service called MyReplayTV. Aided by Web consulting firm Viant, ReplayTV engineers undertook the task of translating to the Web the TV interface of the company's ReplayTV DVR set-top box, creating the first Web site for remotely controlling an in-home consumer device. Once completed, registered MyReplayTV users could log in to add, delete, or modify the shows recorded by their home devices, and search for programs.

Online Advantage

The MyReplayTV project was born amid rapidly escalating competition in the DVR marketplace. In the summer of 2000, ReplayTV began looking for ways to distinguish itself from its main competitor, TiVo. The company decided that extending its service to the Web would be an attractive selling point for ReplayTV's tech-savvy, early-adopter target market.

"The Web site adds additional functionality and value to our service versus the competition," explains ReplayTV product marketing manager Lance O'Hara. "If a customer is comparing the features of ReplayTV to the competition, this is always one area we can promote that the majority of the other competitors can't."

The MyReplayTV development team took the project from concept to launch in about five months, allowing SONICblue to market the service for the winter 2000 holiday season. Among the project's many challenges was the mandate to exploit the benefits of the Web while not radically changing ReplayTV's existing back-end systems or diverging too far from the familiar TV interface.

Hardware Limitations

Beginning in July 2000, the ReplayTV engineering and user groups met to talk about the technical feasibility of the plan, what functionality it could implement, and the technologies that might be used. With nothing remotely like MyReplayTV online at the time, the team had no established model to guide its work.

One limitation that would nag the team throughout the process was the batch-oriented nature of the existing ReplayTV system. The ReplayTV devices' communication with ReplayTV network servers is limited to a nightly dialup to send and receive commands. Thus, requests sent through the Web interface could take up to 24 hours to reach a user's DVR.

"That was a huge trade-off," says MyReplayTV project manager Ed King. "All the information that's sent to your box every night is being put together during the daytime. So when you call up, we send you a batch of preprocessed data. With the time constraint we had, we couldn't reengineer the whole back-end to [work in] real time." The team would have to design around this so-called "blackout period" and make sure users understood the constraint, while leaving open the ability to move to a real-time system as the abilities of new ReplayTV devices evolved.

With a rough concept in hand, ReplayTV hired Viant that August to guide the development process. Viant then determined that the first release should concentrate on making the home devices fully controllable through a desktop Web browser, leaving open the possibility of extending the service to other Web- and voice-enabled devices in the future. With the technical boundaries in place, the Viant and ReplayTV teams would work toward a concrete design plan.

Design Direction

According to Viant design lead John Armitage, the Viant team proposed a user-centered approach for the project, involving test users early in the process to guide design choices. The Viant team members began by becoming ReplayTV users themselves.

"We familiarized ourselves with the unit. That in itself was new to us. DVRs were pretty new at the time," says MyReplayTV project manager Larry Tsai, who worked on the project as a Viant employee. "We familiarized ourselves with that UI and started to sketch out designs for the Web."

Viant also wanted to better understand what role the service would play in the lives of actual users, so with ReplayTV's help, Viant contacted a few DVR beta testers to arrange in-home visits. "We took pictures and videotape of their homes to see how they set up their Replay," Tsai explains. "We looked at how they actually used and interacted with it in their own environment."

Armed with demographic information from ReplayTV and the results of the in-home sessions, Viant developed typical user personae and scenarios. Viant's strategy was to target well-defined user types, rather than designing for an infinitely variable mass of users, which could have led to a bloated and cumbersome product. The service would be slanted toward the tech-savvy power user, and Viant would find ways to design the site to help users outside that group adapt.

User Focused

Viant conducted eight 90-minute one-on-one user experience interviews with a group that was representative of the target personae, including current ReplayTV owners, TiVo owners, single adults, parents, and teens. Each group included both expert and intermediate Web users. The goal was to test the effectiveness and the intuitiveness of the site: Would users be able to understand and use the site right away? Did the Web interface match the TV interface?

Test users were shown hand-drawn, oversized, black-and-white paper prototypes. According to Tsai, these sketches emboldened the testers to be critical and let Viant concentrate on the functional concepts of the site, such as which buttons users pressed and which labels made the most sense. Viant then made quick changes based on user input.

In comparison, Tsai says that, typically, test users for other Viant projects were shown slick HTML mock-ups in a focus group setting much later in the process. The change in procedure was an important part of the user-centric approach used for MyReplayTV. "We wanted to be really interactive with the users and to be able to change things on the fly," Tsai says. "We found this to be a lot more useful—getting immediate feedback, making immediate changes, making it very iterative."

The testing revealed much about the possibilities and limitations of the MyReplayTV concept. Users were unclear about the blackout period, so Viant recommended that MyReplayTV use contextual help items and gray out affected time slots in the online Channel Guide to clarify the constraint. The input of inexperienced users helped Viant refine the positioning of other help elements and tweak the registration and login processes for new users.

Test users perceived MyReplayTV as a research and planning tool. In response to that perception, Viant designers exploited a computer screen's higher resolution to make more information available to MyReplayTV users than can be displayed on a TV screen. The search function also benefits from the Web interface, allowing quicker and more refined searches than are possible with a remote control and TV interface. Armitage says that the Web interface retains the color scheme, layout, and functionality of the TV interface—the good and the bad—for consistency's sake.

Back-End Support

The biggest engineering challenge was translating the data from the legacy system—which supports the TV system—to the Web, Armitage says. MyReplayTV is engineered as a three-tier system that separates presentation, application logic, and data management. The presentation layer uses the Apache Web server on Sun UltraSparcs running Solaris. These handle HTTP requests from MyReplayTV users, serve static pages, and hand off servlet requests to the Tomcat Java application server. According to King, Apache and Tomcat were chosen because they are inexpensive, robust, and well understood.

The data management layer had to support personalization features for both Web users and the set-top box. Because of the complexity involved, ReplayTV chose Oracle 8i, as the engineering team was already familiar with that product.

MyReplayTV engineers chose to first craft a well-defined Java API for the project, so that Viant could use that API's specification to guide its development of the front-end operations. ReplayTV's engineers used the same API to guide their modifications to the existing back-end systems.

The MyReplayTV Java API presents an externally accessible interface to the Web server that hosts the MyReplayTV site. Controller servlets on the front end validate user input as necessary and make HTTP GET requests to retrieve Replay Guide, Channel Guide, or user profile information from the back-end servlets. These servlets handle all data retrieval and check data integrity.

Implementing XML

The data returned by the back-end MyReplayTV servers is formatted as XML. King says that XML's structured-yet-flexible nature made it the best choice for ReplayTV's data communication scheme—especially because it was already being used to transmit program listings and other data between ReplayTV servers and set-top boxes. Client software running on ReplayTV devices uses proprietary C++ code to parse and encode XML data.

For MyReplayTV, King says adding an XML parser to encode Web requests sent to the back-end would have added too much time to the development schedule. Instead, parameters are sent using the standard HTTP GET mechanism. This design allowed for some flexibility in the database schema, which changed often in the development process. It was also easier to test, King adds. Developers replicated the servlet's action during testing by sending the GET arguments from a browser. The servlets return raw XML, which can be displayed by Internet Explorer and other browsers.

After receiving XML data from the ReplayTV back-end, the Web interface must process it and present it in a useable form. Thus, the presentation layer needed an XML parser to handle data returned from the application layer. Viant chose a software package called BreezeFactor for that purpose.

BreezeFactor saved the Viant team time by automating the task of binding the XML data to Java objects. The objects were then used to create Java Beans. The front-end servlets at the application layer use JavaServer Pages (JSP) to describe how the data from each Bean will be displayed. They then output HTML to the Web server at the presentation layer.

Launch And Beyond

"The short time frame of the project forced many of the development steps to happen in parallel, whereas they would otherwise [have happened] sequentially," explains Darius Kasad, Viant's tech lead on the MyReplayTV project. "There were changes to the API and the DTDs [Document Type Definitions of the XML datasets] throughout the course of the project. To mitigate some of the risks of these changes, we wanted to automate the process, rather than spending time writing code to parse XML ourselves and updating that code as the DTDs evolved."

With a system in place to collect and store Web requests in a database, ReplayTV engineers still had to modify the back-end system and the software on the set-top box itself to act on the data. Before MyReplayTV, the set-top box could only handle programming requests sent from a user's remote control.

Version 3.0 of the set-top box software incorporated the necessary changes. The firmware upgrade was rolled out to users during a regular overnight dialup connection in January 2001, to coincide with the launch of the MyReplayTV service. SONICblue now sees regular use of MyReplayTV by about 10 percent of its DVR users.

For Viant, developing MyReplayTV was a valuable learning experience. Armitage says that he used the project as a lab to teach the user-centered design technique to his colleagues for use on future projects. For SONICblue, MyReplayTV strengthened its role as a pioneer of TV-computer convergence. ReplayTV's O'Hara says, "With MyReplayTV, we removed the boundaries and barriers so that customers can program their ReplayTVs to record. That was totally new."

---

Doug is a freelance journalist and Web producer in Austin, Texas. You can contact him at webpro@daddison.com.

RELATED ARTICLES Intel XML Software Suite 1.2 Released Intel, MuleForge Team Up for XML Tools XML as a Content Platform Code Signing in Adobe AIR Dr. Dobb's Agile Newsletter 08/08		TOP 5 ARTICLES No Top Articles.