During the rise of the Internet, the portal's roles as an aggregator of content, a means of personalizing Web access, and a conduit to various online communities became a mainstay of life on the Web. Accordingly, portal sites enjoyed the position of "sure thing" players because they delivered value in traditional media terms: an audience that was easily measured in terms of hits, page views, and memberships. But today, "portal" has become an overloaded term, implying everything from network access to the familiar stock quotes, news, and weather aggregation services.

However, a second generation of portals will soon bring a new focus to the portal concept, one with enduring value and a solid economic foundation. In the wake of the great dotcom implosion, a new set of metrics is governing the growth of the Web. Based on "old economy" terms, it focuses on practical items such as cost of transaction, cost of services, brand loyalty, and customer satisfaction. The next generation of portals will become the prime delivery vehicle for a new set of network services that satisfy these metrics. These new services are generically termed Web services, and will become the dominant technical and business model for the next generation of networked applications, including those that drive Web sites.

In the broadest sense, Web services refers to networked access to highly focused application functions, using XML messages in a loosely coupled fashion. Web services include a number of attributes that offer a vastly expanded range of possibilities compared to the present model based on tightly coupled interaction inside the firewall. Web services are highly flexible in terms of connections, offer deep granularity in application functions (as opposed to today's monolithic packages), and fully support the construction of systems that become a "network of networks." Taken together, these attributes support the genesis of second-generation portals with significantly expanded capabilities that will make them the prime delivery vehicle for the coming wave of Web services.

The move from relatively isolated, large-scale applications to finer-grained Web services will have two major impacts on portal servers. First, Web services require a multiplicity of portal delivery channels to any number of devices, ranging from PCs to LCD displays in automobiles. Second, careful architecture and application design with portal delivery channels in mind will create "services on demand" which offer significant benefits to both businesses and consumers, such as "anywhere, anytime" access.

Portals
The integration of Web services into portal environments will represent a major shift in how a portal is ultimately defined. The first generation of portals, which still dominates major Web sites across the Internet, can be thought of as a combination of Yellow Pages and a clipping service. As a Yellow Pages directory, it simply offers a convenient method of navigating to sites of common interest for the portal's user population, functioning as a switchboard to connect individual clients to a community of services. As a clipping service, it processes streams of data such as stock quotes or news headlines, and formats this influx into HTML for viewing on client browsers. Both of these functions are based on delivery of relatively shallow content - clicking on a news story merely displays the story but has no other relationship to applications within the enterprise. Portals haven't been used to foster user interaction in a way that substantially enhances the user's productivity, especially in commerce environments.

In the second generation of portals, what was once static becomes largely dynamic. The portal's data channels are no longer passive conduits for an influx of data. Instead, they become interfaces to programs performing active functions on behalf of the user. A content channel in a portal can be thought of as an active content generator, driven by user or external events. Designing applications for this content-oriented delivery mechanism redistributes the processing load in a way that allows data to be delivered in highly specific, value-added ways to the end user. The portal becomes the underlying application's API consumer, intermediating between users and their collection of applications so that aggregation or process-driven application coordination become the typical user work flows.

This new order allows the user to interact in a way that provides highly customized information. In effect, the portal is leveraging the API to create a dynamic content channel. Each channel is an object representing the application that feeds the channel. Portals can now be developed so that they become collections of such objects arranged to maximize the user experience.

To provide these dynamic content channels, second-generation portals encourage the shift of processing to the server side and continue the trend toward "thin" clients. Rather than thick clients that perform editing, local logic functions, or presentation management, those functions are handled by the server-side business logic or the multi-channel delivery modes of the portal server itself. As a return on this investment in new architectures, software developers will have greatly expanded flexibility and control when developing transactional applications.

For example, consider an application that performs transactional updates of several systems and has been targeted at a wireless handset for consumer use. In a browser-centric world, or a desktop thick-client model, it's tempting to use a heavier-weight client or client-side applet to perform some of the transaction sequencing. Using a second-generation transactional portal, the work of ordering the transactions, handling failures and recovery, and providing updates to the user is handled entirely on the server-side. Factor in a lower reliability client-side environment such as a wireless handset, and the benefits of refactoring the business logic with a heavy emphasis on the server side are clear; the portal server provides device independence when the heavy lifting is safely on the well-managed, predictable side of the data center.

This new breed of portal provides an ideal solution in the current e-commerce milieu, where the demand for portal technology is firmly established. Although portals first gained a foothold in the business-to-consumer space, where they were the province of large-scale ISPs, there's now a clear recognition that portal technology will play a critical role in enterprise infrastructures, where they'll help employees make sense of an overwhelming amount of data from both within and without the corporate firewall.

This compelling problem of data overload has already caused the market for Virtual Private Networks to grow to more than $1 billion in 2001 as enterprises attempt to develop expensive, customized methods of giving employees "anywhere, anytime" access to corporate information.

In contrast, second-generation portals will give enterprises an off-the-shelf-approach to this problem that promises to be highly cost-effective. A portal server - and its surrounding infrastructure - will provide a highly flexible template that can be quickly adapted to specific corporate needs. A host of industry standards, such as the J2EE platform, XML, SOAP, and many others, help to make this template universal in nature. Thus, the primary growth market for transactional portals is likely to be within the enterprise, where cost savings and improved decision-making are enabled by providing employees, partners, and customers with direct access to the right combinations of applications.

New Flexibility
As transactional portals grow in number and scope, they enable further flexibility in the number and types of applications developed, and expand the scope of interactions these new applications have with existing internal and external systems. Logical connections between systems will be driven more and more by user demands, as opposed to the dictates of tightly coupled applications. This growing diversity of highly specialized applications and the density of their interactions will require a new form of control and coordination that shields the user from all the underlying complexity it produces. The solution will be a highly evolved form of Web services called services-on-demand, which continuously monitor the user's current situation in terms of device currently in use, geographical location, and notification preferences.

This new type of Web service remains continually aware of all aspects of the user's present context, both in a formal and personal way. For instance, a presence-enabled service knows who the user is, his or her present role and preferences, privacy requirements and security level, past history in terms of required service, adherence to service agreements, geographical location, and type of device currently in use. In more mundane terms, the service recognizes that if you're in a meeting with your cell phone turned off, an SMS message won't be nearly as effective as a text page, but both are preferable to an instant message if you're not online at the time.

How will these services-on-demand be integrated into second-generation portals? To recap, the dominant feature of the new portals themselves will be the inclusion of dynamic data channels that coordinate activity between user applications. To make these channels "smart" in the sense described above will require an entire architecture where the portal presentation aspect is coupled with Web and application servers, as well as functional blocks driving policy, network identity, and service management. Also required will be provisions for automated workflow that couples with back-end databases and legacy systems, in addition to other Web services.

The dynamics of this architecture are best understood by following the flow of data on one of the portal's dynamic channels. When a service request comes in over the portal, it includes the user's present context in terms of location, current access device, role, and even time of day. This information is captured and fed into the application server, the central engine where the Web service executes. Here the service calls upon other sections of the architecture to examine and process parameters such as current context, policies regarding the present request, and applicable business rules. This includes orchestrating the service's overall workflow and coordinating tasks such as accessing databases, legacy apps, and other services. It also includes consulting an entire range of policy microservices covering items such as identity, security, privacy agreements, and specific user preferences and history. At the same time, another series of microservices cover management tasks, such as billing and enforcing quality of service levels. When the service is prepared to output a response, it uses the portal to generate the response in the proper context, such as reformatting for wireless interfaces and personalizing it for a specific user.

The next wave of growth on the Internet will be fueled by business models that rely on decreased cost or new sources of revenue; services-on-demand will be a major driver of that growth because the fundamental architectural goals are to connect applications and data assets to the right communities of interest at the right time. With this type of architecture in place, it becomes clear that the next generation of portals is poised to deliver a whole new range of services-on-demand capable of adding significant value to those businesses and individuals that use them.

Author Bio:
Hal Stern is the CTO of iPlanet. In that role, he identifies and leads adoption of distributed technologies, aligns iPlanet and Sun Microsystems engineering functions, and maps high-level vision into practical technology approaches. He was the first employee from Sun's field organization to be promoted to Distinguished Systems Engineer. Stern is the coauthor of Blueprints for High Availability. hal.stern@sun.com