In response to lutfi at smartconsultant dot us :

(see my post on
http://fr2.php.net/manual/en/
function.dom-domdocument-getelementsbytagname.php
)

Use this class I wrote:

class XPathableNode extends DOMNode
{
    protected $Node;
    protected $DOMDocument_from_node;
    protected $DOMXpath_for_node;

    public function __construct(/* DOMNode */ $node)
    {
        $this->Node=$node;
        $this->DOMDocument_from_node=new 
                                                          DomDocument();
        $domNode=$this->DOMDocument_from_node
                                  ->importNode($this->Node, true);
        $this->DOMDocument_from_node
                                 ->appendChild($domNode);
        $this->DomXpath_for_node = 
                    new Domxpath($this->
                                         DOMDocument_from_node);
    }

    public function convertHTML()
    {    return $this->DOMDocument_from_node
                                                             ->saveHTML();
    }

    public /*DomNodeList*/ function applyXpath($xpath)
    {    return $this->DomXpath_for_node
                                                       ->query($xpath);
    }
}

(sorry for the display... What a terrible hinderance on the
part of php.net !)

Then :
Make a new XPathableNode out of your parent node.
You may then retrieve a DOMNodeList from it by applying a
xpath (thus being able to specify the depth  and name of 
elements you want).

Has got me around some (of the many) DOM awkwardnesses a few times.

;o)