The World Wide Web (WWW) and Collaborative Software

The WWW has spurred interest in network based collaboration. It is important in starting out to comment on the WWW phenomenon. As time goes on, WWW applications will become more powerful and more ubiquitous -- this much seems clear. At the same time, most WWW applications are based on a rather limited set of protocols. At heart, WWW services assume a client-server application using http and html. Surely, ftp, smtp, gopher, etc are woven in as additional services, and http is vastly expanded in capability by CGI and most recently Java, but at heart, the WWW begins with http and html. We suggest that these protocols are optimized for very general publishing functions -- for information dissemination. Figure shows a simple graphic of the http protocol. As can be seen, a client sends a URL to a server, which looks up the URL in its file space and returns a byte stream representing the object to the client, or returns a message that the object does not exist. The object, as far as tserver is concerned is simply a byte stream. The client interprets the byte stream for display either as a series of lines of ascii text, or as an HTML document. There is no provision, under this basic form, for the client to send anything but a request, or for the server to do anything but return an entire object. As soon as the transaction is complete, both parties disconnect -- no state information is maintained. In addition, the nature of the request at this level is such that information requested is either available to all who ask for it or to none. There is no ability at this level to discriminate between requestors.

At the next level, which emerges over a series of years, http grows up. This occurs in a variety of ways as described below and shown in Figure :

• Clients are made more sophisticated in terms of the complexity of the HTML document they can manage. Forms, mail, image handling are all enhanced. Ultimately, this should lead to the ability to handle documents of types other than HTML. Some readers will understand that HTML is two things: first, it is an SGML compliant DTD, and second, it is an extension of SGML to allow for a primitive but very effective form of hypertext.

• Servers are made more sophisticated by the addition of new methods. This comes in two forms. First, a simple table lookup is added that allows image coordinates to be passed back to the server. These coordinates are resolved by checking an imagemap file (Table) that selects a URL to be passed back based on image areas associated with URLs. If the coordinates fall within an area specified in the image map, the associated URL is returned. The second, much more general refinement of server capability occurred when a new protocol was associated with the forms capability of the client. In this case, the URL passed to the server could identify an executable program and the URL could contain data to be passed to the executable that would allow conditional response to the URL. The executable program is constructed, located, and run in accord with what is known as the Common Gateway Interface(CGI). At the simplest level, the program could select different pages to send back based on the data. In practice, the program most frequently uses the data to generate information from a database which is then constructed on the fly in the form of an html page that may contain additional URL's that access the Common Gateway Interface.

• Finally, Java is introduced which provides on the client the same kind of data processing functionality CGI gives the server. A full explanation of Java requires more time than we have here, but it provides a means of delivering a "document" from the server to the client that can then be safely run on the client providing conditional activity on the client that will generate differential requests to the server and that will interact locally with the user at the client level in a conditional and sophisticated manner. It is also the case that "cookies" are introduced, greatly overshadowed by Java, that provide a mechanism for maintaining state information about client server relationships.

  
Figure: The Basic HTTP/HTML Paradigm

  
Figure: HTTP/HTML enhancements via Forms, CGI, and Java

In this process, http is made into a vastly more powerful and sophisticated protocol. At the same time, it remains at the core a protocol intended to provide a transaction oriented relationship between a client and a server optimized for the purpose of information dissemination. While it is now possible for clients to deliver information back to the server to chnage the state of the server, this is far from the core functionality provided by the system. We will suggest that for network based authoring, additonal functionality will be required which suggests protocols other than http. At the same time, it would be a mistake for us or anyone else to underestimate the momentum behind the WWW, http, and HTML.