I don't see the problem, whether the markup in the XML document is
<foo></foo> or <foo/> or <foo />, in the DOM object model or in the
XPath data model you will have an element node with name foo and no
child elements so you can simply compare on the node type, on the node
name and if needed on the child nodes.

Well you could use the XPath data model which only knows text nodes and
does not distinguish between text nodes and CDATA section nodes. Or your
comparison code would need to ensure that text nodes and cdata section
nodes are compared based on the node value, that will be the same then
whether you have e.g.
  <foo>a &amp; b</foo>
or
  <foo><![CDATA[a & b]]></foo>
does not matter, you will still have the same node value.

As for line endings, with .NET 2.0 and later, if you use
XmlReader.Create to parse your XML, then line endings are normalized
anyway as the XML specification requires. Thus if you work with
System.Xml.XmlDocument then use e.g.
  XmlDocument doc1 = new XmlDocument();
  doc1.Load(XmlReader.Create("file1.xml"));
instead of
  XmlDocument doc1 = new XmlDocument();
  doc1.Load("file1.xml");
then line endings are normalized and your comparison code does not have
to worry about that.