I have a quite stupid question. How can I make sure that my XML mixed content node doesn't get mixed up? I have, say, an XML structure resembling this.

<root>
 <book>
  <title>Stuff</title>
  <description> This book is <i>great</i> if you need to know about stuff.
                I suggest <link ref="Things">this one</link> if you need to know
                about things. </description>
 </book>
 [other books]
</root>

I need the final content to look like this

<h1>List of books</h1>
<h2><a name="Stuff"/>Stuff</h2>
<p> This book is <i>great</i> if you need to know about stuff.
    I suggest <a href="#Things">this one</a> if you need to know
    about things. </p>

But I can't extract the portions of the text node, I always grab the whole thing. I'm using the descendant axis. Any clue what I'm doing wrong?

Here is my xslt:

<xsl:template match="description/*">
    <xsl:for-each select="following-sibling::*">
            <xsl:choose>
            <xsl:when test="name(.)='link'">
                <a href="{@ref}"><xsl:value-of select="."/></a>
            </xsl:when>
            <xsl:when test="name(.)='em'">
                <em><xsl:value-of select="."/></em>
            </xsl:when>
            <xsl:otherwise><p><xsl:value-of select="."/></p></xsl:otherwise>    
        </xsl:choose>
    </xsl:for-each>
  </xsl:template>

Please note that the enclosed XML and resulting html are merely examples, I have to deal with a bigger structure which I'm not enclosing in, for sake of clarity.

xml xslt axis
Edit
Report