Building this website

Okay, so my website doesn't look like much. But I've had it up and running for almost eight years now. Over time HTML changed quite a bit, and I keep experimenting with new browser technology. Some documents have been on my website since 1995 though. They've seen many layouts come and go.

XSL Transformations

Each time I want to change the layout of the page, I don't want to create a new HTML page and copy-paste the content into that page. No, instead I use a source webpage and translate that page in the final webpage with the right layout. Here's my setup:

• All content is stored in XHTML (I removed all styling from the HTML, and then converted it to XHTML on a rainy sunday). I like using plain (X)HTML for my content, since I still have lots of styling freedom in the form of bulleted lists, headings, tables, preformatted text, links, images of all sorts, etcetera. Another advantage of writing in plain HTML is that I can use class attributes on the HTML elements, and assign styling to those classes in the CSS stylesheet.
• There's an XSL stylesheet that describes the target HTML layout, named note.xsl.
• I use libxslt by Daniel Veillard (specifically the Windows build by Igor Zlatković) for the transformation from content XHTML to final XHTML. There's a command line utility named xsltproc included with this library to take an XSL and XML document and transform the XML in to another XML document.

Makefile

To generate all the final HTML pages in one go, I use GNU make. I wrote a Makefile with a static pattern rule to create .html files from .xml files. As you can guess, all filenames of my content files end in .xml. This webpage for example has a source file named computing.xml.

Make checks to see if there is a final webpage, e.g. computing.html. If so, it compares the timestamps of both prerequisite content document and target final webpage to see if the final webpage is up-to-date with the prerequisite content document, e.g. computing.xml. If there is no target webpage, or the webpage is out-of-sync with the content document, make will run the XSL transformation:

pages = $(patsubst %.xml,%.html,$(wildcard *.xml))

all: $(pages)

$(pages): %.html: %.xml webpage.xsl
    xsltproc --output $@ webpage.xsl $< 

A simple `make all` in the Bash shell generates all my webpages. Note that you have to put a tab, not spaces, in front of xsltproc! And also notice that webpage.xsl is a prerequisite, which means that if I edit the XSL, all pages will be regenerated.

Backup

As an advanced form of file backup I use version management on all my website files. I've been running Subversion (SVN) for half a year now and I'm quite content with it:

• Excellent documentation!
• Powerful properties system
• Major plus for websites: it automatically recognizes that it's dealing with images, something you have to teach CVS to do for you.
• Directory source control, also very welcome for websites
• Familiar to CVS users

Oh, if you didn't understand by now, this is all command line stuff. I use the cygwin environment on Windows, so that I can use the same tools I would use on Linux.