as a child of a <tag>mediaobject</tag> to specify a
directory into which any long-description files for that
<tag>mediaobject</tag> will be written.</para>
+
+<para>The output directory specification is inherited by all
+chunks of the descendants of the element. If descendants need
+to go to a different directory, then add another
+<tag class="xmlpi">dbhtml dir</tag> processing
+instruction as a child of the source element
+for that chunk, and specify the path relative to the
+ancestor path.</para>
+
+<para>For example, to put most chunk files into
+<filename class="directory">shared</filename>
+but one chapter into
+<filename class="directory">exception</filename>
+at the same level, use:</para>
+
+<programlisting><![CDATA[<book>
+ <?dbhtml dir="shared"?>
+ ...
+ <chapter>
+ <?dbhtml dir="../exception"?>
+ </chapter>
+</book>
+]]></programlisting>
+
+
</refdescription>
<refsynopsisdiv>
<synopsis><tag class="xmlpi">dbhtml dir="<replaceable>path</replaceable>"</tag></synopsis>
<doc:pi name="dbhtml_filename" xmlns="">
<refpurpose>Specifies a filename for a chunk</refpurpose>
<refdescription>
- <para>When chunking output, use the <tag class="xmlpi">dbhtml filename</tag>
+<para>When chunking output, use the <tag class="xmlpi">dbhtml filename</tag>
PI as a child of a chunk source to specify a filename for
- the output file for that chunk.</para>
+ the output file for that chunk. Include the filename suffix.</para>
+
+<para>You cannot include a directory path in the filename value,
+or your links may not work. Add a
+<tag class="xmlpi">dbhtml dir</tag> processing instruction
+to specify the output directory. You can also combine the two
+specifications in one processing instruction:
+<tag class="xmlpi">dbhtml dir="mydir" filename="myfile.html"</tag>.</para>
+
</refdescription>
<refsynopsisdiv>
<synopsis><tag class="xmlpi">dbhtml filename="<replaceable>filename</replaceable>"</tag></synopsis>