Networking

How Applications Work

Traditional end user applications generally have two major limitations. They limit the user's ability to reformulate or restructure the interaction, and they interact with data from a pre-structured, single database. For complex solutions, like Human Resources, Help Desk or Sales Force Automation, the data required by the user often resides in multiple legacy databases and a large portion is semistructured, idea-based information, not the discrete quantities and numbers that databases manage so well.

For these reasons, most application packages in these markets focus on the logic that supports the structured part of the problem and on the user interface. The legacy information may be acknowledged, but is not the main thrust of the marketing and sales effort. The obligatory statement that, of course, the legacy databases can be integrated is made, and the focus returns to the structured process logic. And yet, in the implementation, the knowledge-base problem is the significant barrier to an effective application, not the structured process logic.

True Intranet applications are still in their infancy, but one point should be clear. An Intranet application starts from the perspective of the knowledge base, not the perspective of the structured application logic. The application logic is one resource in the knowledge base that leverages the environment to perform specific functions. This is a significant difference that is possible only because of the vendor-independent communication and content standards of an Intranet.

We can expect Intranet applications to evolve over time. The applications we see today tend to be traditional applications, modified to take advantage of the standard user interface. This type of application retains the tight integration of functionality in a proprietary implementation and the dependence on its database for information discovery and display. The next wave of Intranet applications will begin to integrate solutions that extend to functionality beyond the structured and semi-structured processes of traditional applications. The use of spider-based discovery agents will begin to feed application logic, and the logic will begin to facilitate the interaction of higher level ideas over the raw manipulation of data that characterizes computer applications today. As our knowledge about discovery agents and their implementation improve, the application logic will begin to unbundle from today's structured database model. This will both increase the need for vendor-independent, object-interface standards and facilitate the unbundling of integrated logic into functional objects. See Corel's Java for Office as an example of early moves in the direction of unbundling logic into functional objects.

So what will these applications look like? They will be built on the standard content and discovery agent model. They will focus on semi-structured and unstructured parts of the problem. They will facilitate self-customization by the user. Structured logic and processes will be developed and shared by anyone, and many will be single-use "throw-aways."

The one characteristic of an Intranet that makes it different from all previous computer-based infrastructures is not the wealth of information available, but its ability to make everyone a publisher (and soon a programmer). This fact is often either overlooked or viewed as a problem that has to be managed. In fact, if a medium does not allow everyone to publish, outside the "fill-in-the-blank" structure of previous computer technology, then real communication cannot happen. The organizational interaction is limited to dictates and highly structured feedback.

How can we move to the much desired "learning organizations" if we hide in structure and cannot embrace some chaos and inefficiency? If we already know how to structure the problem and the information, how can we learn? Learning is the process of discovering structure. If we can only manage the knowledge (the repository of our previous learning) that fits into our currently understood structures, how can we advance our knowledge as we learn radically new things?

This is not to say there is not a place for structured processes or broadcast information. However, the new infrastructure does allow us more options in the way we approach and define problems. The key considerations in Intranet application architecture will be twofold: the desirable level of structuring and when content needs to be pushed rather than pulled. A new generation of "push" tools are emerging that give users more control. See Marimba's Castinet product, the Pointcast Network product or FirstFloor's Smart Delivery product as examples. Complex Intranet applications will support a mixture of push and pull possibilities, applied to gain the best overall effectiveness.

In their book, Decision Support Systems, Peter Keen and Michael Scott-Morton identified three classes of processes which they called, structured, semi-structured and unstructured. Prior to the time of their work, computer applications primarily focused on structured processes. These well-understood, repeatable processes fit the original "batch" mode of the technology quite well. The concept of Decision Support Systems opened the world of computer applications to the support of semi-structured processes. This was enabled by the advent of interactive computing from video terminals, which meant that users could interact with programs while they were running.

The introduction of Intranet applications extends computer functionality to include support for unstructured processes. The user has the capability to scan (browse) and screen (search) unstructured information to help formulate more specific questions or to stimulate new ideas or approaches. The process building blocks in the previous section can be combined in various ways to support all three process classes. Additionally, the technology allows the application author to mix modes of support within a single solution.