Information is renewed at such speed on the Internet that it is humanly impossible to follow the evolution of a competitorís offer, the news of an economic sector, or the appearance of technological innovations, without the help of electronic assistants. We have seen that in certain circumstances, information retrieval can be done without agents, and can even be optimized manually. This is not possible for advanced monitoring activities. Even if you obliged a person to remain "connected" to the Internet night and day to monitor just 10 large sites, you would still not obtain satisfactory results. The quantity of information is too great for a human to be able to detect all the changes that take place. Therefore, he must "collaborate" with monitoring agents. Schematically we can distinguish two major agent families; "push" and "pull" type agents.
When you go on Yahoo to undertake a search, it can be said that you are "pulling" information from the Yahoo server to your machine. Effectively, you have taken the active step of starting your computer, connecting to the Internet, going to the Yahoo site, providing the keywords for your search, and finally downloading the documents to your computer. You are doing pure pull... However, you can continue to pull by delegating certain repetitive tasks to an agent. If for example, instead of connecting directly to Yahoo, you delegate this task to your search agent (for example WebSeeker), it is still pull. In this case, the agent replaces you and it "pulls" the requested information towards you. In the pull approach, what is important is that you are active in your choices, even if you entrust the repetitive work to the agent. In the same way, if you decide to monitor the Web site of a competitor with a monitoring agent such as Webspector it is still a pull even if the agent stores the pages that it finds for you, and it gives you the impression of "pushing" them to you.
With push, on the other hand, you play a much less active role because information is "pushed" to you. It is exactly the same principle as television. You have many channels that disseminate information, and your role consists only of choosing the channel that is appropriate to you, and then following it. On the Internet it is the same thing. You have software of the push type (PointCast, Marimba, BackWeb, etc.), which gives you access to hundreds of information channels (CNN, New York Times, Washington Post, etc.), often with a set of themes (weather, sport, markets, financial, etc). Push software exploded towards the end of 1996 making a lot of noise about itself. Some affirmed that from now on, the whole Web was going to be transformed into a "giant television" in which everyone would disseminate information from their own channel. In reality, it is the enormous financial stakes (in particular related to advertising) that caused such enthusiasm. The information obtained by these web-casting (or Web-diffusion) agents is completely uniform and standardized, and often adapts to a mass diffusion. It is therefore very useful for following general news, but completely insufficient for undertaking personalized monitoring activities. With push technologies, it is the manager of each channel who decides its content, and who diffuses it (through a server), whereas with a pull agent, if you decide to create a channel, you are the master of its content and it is your agent (a software client) that downloads the information that corresponds to your choices. Human intervention and personalization are much greater with agents of the pull type.
Sometimes the distinction between pull agents and push agents becomes very subtle, especially since the most recent browsers (starting from versions 4.0 of Netscape Navigator and Internet Explorer) have integrated the two technologies simultaneously. In addition, the fact that the official channels can co-habit with personalized channels will tend to reinforce this confusion.
We are persuaded that the two systems will co-exist more or less peacefully, simply because they are complimentary. To think that the Internet will be transformed into a single system of Web diffusion is like thinking that television is the only media which will survive, and that from now on nobody will make a choice and buy a daily newspaper or go to a library to look for particular information... In addition, even if the push model succeeds, and if the Web is transformed little by little into an enormous television with a multitude of official and non-official channels, we will find ourselves with exactly the same information problems on the Internet as today (hyper abundance of channels which are not always relevant or reliable...).