Most of the information available today through push/pull technology consists of snapshots of subject material. Over the next few years, however, I believe that this technology will change from being just a provider of "short information bursts" to becoming a central element of business -- particularly electronic commerce -- applications. The technical underpinning of push/pull is actually a type of middleware that allows for discrimination based on very specific criteria and user requirements. Although the middleware behind the movement to push/pull technology has been around for decades, it is just emerging into prominence.
In the past, information was pushed or pulled into systems based on traditional mainframe transaction processing rules with two-phase commit and guaranteed delivery. Typically, when an order was placed, a set of events would be triggered and an acknowledgment would be forwarded to the buyer. These transactions were orderly and secure. What is different today is that we are talking about a less rigid definition of transaction processing -- an uneven volume of transactions without the two-phase commit and security. In the future, providers of push/pull transaction software are going to have to add some of the security and guarantees from traditional transaction processing. This will be especially important as the variety of information sources involved increases (driven by the Internet and emerging electronic commerce providers) and as systems and software become more globally distributed. In addition, the fact that the Internet itself offers a seemingly unending source of information is forcing organizations to look for techniques to manage the information flow. Clearly all of these issues are related. Deploying transaction-oriented software is one of the key techniques that will turn the wildly uncontrollable Internet into a useful business tool.
Before I discuss the emerging technology further, let's set some definitions:
How has push/pull technology been used? As I mentioned, in traditional host systems, push/pull technology was found primarily in transaction processing systems. Companies such as TIBCO Inc. and BEA Systems Inc. are veterans in this market. Their agent-based technologies allowed transaction processing to be triggered by things such as a new order or the end of a quarter. The benefit that these technologies offered was that they worked well in the structured and orderly world of transaction processing systems. Financial services companies, for example, have been using push/pull technology to track stock information for at least the past 15 years. In fact, most of the commercial implementations of push/pull technology are focused on the financial services industry.
In a commercial software context, push/pull technology is all about managing transactional information. This is a familiar problem. When client/server first hit, I became aware of a new and important trend: All systems were becoming transaction-oriented. Why? Because applications had become more graphical and therefore more accessible, managers started asking for more information from these systems. The ability to integrate information so that managers could use it to reach conclusions became key. This was rapidly followed by the concept of closer integration between customers and suppliers. All of these trends made software that managed and routed information more and more important. With client/server, partitioning and distribution of applications quickly became a reality. As an example, when an application is split across several physical systems with data even more dispersed, transactional agent software is the only way to bring pieces of software together. Again, this caused transaction-oriented middleware to flourish (although without much fanfare).
As I have noted in other columns, the Internet has begun to evolve into the de facto infrastructure for distributed computing, or the Hyper-Tier. What is driving much of the current interest in push/pull technology is the demand to enable people both to publish (for example, push) information to a distributed platform and to subscribe to (for example, pull) specific information a user might need via the Internet. Push/pull technology is based on the idea of allowing agent technology to be programmed so that either the recipient of information or the provider of information can initiate information dissemination. Why is this important? Simply put, in order for individuals to get real value out of the untamed frontier of the Internet and its myriad of data, it is important to set up a mechanism to deal with the data in an organized manner. For example, there are hundreds of places on the Internet you can go for news. Are all types of news equally important to you? Probably not. Both recipients and providers want to be able to ensure that users get what they need and want -- with the same guarantees and security that they had with traditional transactions.
The earliest experience that pioneers had with push technology -- PointCast -- left a bad taste in the mouths of many IT managers as they experienced bandwidth and productivity issues. Stored, redundant data filled networks and disks. Many banned this and like products from the PCs. To respond, PointCast offered an alternative approach that enabled an organization to put one copy of the PointCast infoserver on the server and distribute it internally. Once this was in place, an IT manager could restrict how much and how often information could be refreshed.
In most organizations (other than financial services) now, however, push/pull technology is not something that IT managers will see directly. Rather, push/pull products will be embedded in everything from CNN and Reuters to Netscape and Microsoft foundation software. In fact, both Netscape Communications Corp. and Microsoft Corp. have released betas of their next versions that include forms of push/pull technology.
These publish and subscribe technologies will continue to evolve dramatically over the next three years. Corporations will increasingly depend on the transaction processing technology to manage the flow and quantity of business requests. Users will want to define and customize their environment. Your customers will begin to look at more and more software as though it was a personal application designed to match their needs. Moreover, the push/pull agent-based technology will begin to learn user profiles and understand user preferences. In addition to the overt requests users make in their customer profiles, this software will learn about them by what they ask for and how they buy.
Therefore, when you are evaluating products that employ push/pull technology for your organizations, begin to understand the value of one-to-one marketing in the context of software. Remember that the ability to customize software is invaluable. Begin to implement your distributed environments in ways that provide this value and that focus on leveraging the many years of complex business systems you have built.
Marimba Inc. Marimba's Castanet Transmitter runs on a network server and manages the distribution and maintenance of applications or content ("channel") across the Internet or intranet. On the client side, Castanet Tuner receives and manages channels designated by the user from Transmitter. Channel types include Java applications, Java applets, and HTML. Marimba can be used for software deployment.
BackWeb Technologies Inc. BackWeb embeds a piece of client code into server-based applications so that they can be updated with information based on the user profile. Information may be in any content format or software type, including Java, HTML, and audio or video files. As a result, BackWeb is being used in some cases for software distribution.
Wayfarer Communications Inc. Wayfarer's Incisa delivers "Headlinks," a combination of headlines and links, to users' desks. Incisa can be connected to an organization's intranet to combine employee-specific internal information with information sources on the Internet. Incisa is positioned as an intranet solution with push technology.
TIBCO Inc. TIBCO, The Information Bus Co., offers its development tool, Rendezvous, to create middleware that incorporates push technology within distributed applications. Rendezvous also may send messages through the network only once, unlike the point-to-point philosophy of most push products. Rendezvous includes APIs for Java and ActiveX and supports CORBA.
Microsoft Corp. Microsoft is working with BackWeb and PointCast to incorporate the ability to send news, stock quotes, and other personalized information to users of its popular browsing software, Internet Explorer. Microsoft will use CDF in Internet Explorer 4.0 to distribute the information, or channels. Explorer monitors designated channels, including Web sites, for new information and automatically retrieves the content for viewing.
Netscape Communications Corp. Netscape is working with Marimba to add Netcaster, a push component, to Netscape Communicator. Based on existing open Internet standards, Netcaster lets content providers create channels that may be directed to users of Netscape's client/browser software. Lotus Development Corp. Not to be left behind, Lotus announced its intentions to include a window for receiving content in its upcoming version 5.0 of Lotus Notes. This lets users avoid adding an application to the client merely to accept content. No word is available yet on a ship date.