XSLT processors generally use SAX and DOM as the core parser and RAM organisation method respectively.

SAX is a very low-level XML-to-event-chain parser, and DOM is just a list of interconnected nodes.

XSLT is a definition language that can sit on top of that, controlling the process of turning an input DOM into an output DOM.

Anything you write in XSLT could be implemented using SAX and DOM, but a single-input-file-to-single-output-file transformation (or even a multi-input-file-to-single-output-file transformation) is a relatively simple XSLT process whatever the rules for it are, and you should use XSLT.

<xsl:apply-templates> and mode will be your best friends =P

I'd advise learning the XPath rules for node selection (http://www.w3.org/TR/xpath)

and getting yourself a good XSL processor (http://xalan.apache.org/)