Meeting the WWW Challenge -- An Example

At the ADN, our immediate challenge is to bring new information sources -- and to move old ones -- into WWW. But we want to keep as much useful information as possible in our Gopher server. Thus we need to produce several different formats of a single document -- plain text for Gopher, HTML for the Web, and versions for printing.

The only sensible way to make sure the information in all the different formats is the same is to maintain the document in just one "archival" format, and to create programs to transform the archival format into the various display formats. Plain text is not a suitable archival format because there is no markup, and therefore no indication of how to transform it. Interestingly, even though HTML is a form of markup, HTML is not suitable, either. HTML markup itself will change -- an HTML+ protocol is already on the horizon, and you can be sure that won't be the end of it. HTML also tends to be display oriented, and does not provide for customization for classes of documents.

Instead, we have defined SGML (Standard Generalized Markup Language) layouts for our documents, with the possibility of different formats for different kinds of documents. Along with the SGML definitions, we will develop rules that allow the archival format to be changed into HTML or plain text.

As an example, we have a pilot project to document the available software on tigger, our new UNIX machine. Programmers who install programs on tigger must fill out an electronic SGML form that includes various information about the software: version number, where it came from, a brief description, who knows enough about it to fix it, where to find out more about it, and so on.

These files are processed in two ways. First, we generate an HTML file for each package that includes only selected parts of the information: the name, description, and where to go for more information. An HTML file with pointers to each of the package HTML files is available for anyone to peruse (enter softlist on tigger, or use the URL http://tigger.uic.eduhttp://tigger.uic.edu/softprod/tigger.html). We also process the archival files a second time, to produce another (more complete) set of HTML files for our programmers.

Since we have recorded all the information we think is relevant in the archival files, if the definition of HTML changes, all we have to do is change our conversion program -- not the original data in the archival files. In short, the information itself is stored in a form that is easy for automatic programs to read, yet is independent of any display technology. This will allow us to always use the best display technology available, changing only our conversion programs when necessary.