I use a Java-based tool for directing the transformation of XML to XHTML, and the XSL transformation is applied by a Java XSL tool. However, there are no dependencies with Java for the XSL transformations--I am only using standard XSL. In principle, I should be able to point a browser to the XML and have the XSL transformation take place in the browser, but at least one attempt to do so failed, and so I haven't bothered with it.

So, starting with some XML files, I use a command-line tool called Ant to call the XSL transformation on each on in turn, and dump the output HTML files into a parallel directory structure. Ant is an open source tool typically used for directing a project's build process. It can, for example, only work upon those files that have changed since the last "build". The XSL transformation uses the built-in, default XSL processor bundled with Ant.

There is no magic to this part of the process. However, there are some extra transformations outside of the pages themselves. There is one list of all project files, in XML format. For each file, only the file name is given, as

<file src="local/xml/technote/xml_and_xsl.xml" />

Last, there is one more file maintained manually, namely an XML file listing topics on the web site. This is used to build the high-level navigation links for the site. Currently the topic support is shallow--no nested topics. Each topic can have a short and long name (short used for navigation links), key, and description. The topic list is also used as a lookup (to get the full topic name), and will be used to generate a site map later on.

That's the basic gist of how this new system works. I realize there are many tools in available to do this sort of thing. But working on it allowed me to brush up my XSL, and I also have full control over my very simple XML system. The entire effort of negotiating the XML and XSL, and not including writing the content and mucking with CSS, probably took three days. It seems, so far, like a worthwhile effort.