$a << $b is true when the singleton node in the variable "a" is
before the singleton node "b" in document order in the same tree.

$a >> $b is true when $a is after $b in document order in the same
tree.

Now I said "in the same tree", but it works it just is implementation
dependent when comparing nodes from one tree to another tree.  One
processor will have it one way and another processor will have it in
another order.  So you will get true/false when comparing document
order between nodes from different trees, but not necessarily the
same answer when using two different processors on the same tree.

The spec does not seem to be talking about nodes from different
documents. Is this a deliberate omission?

What's the rationale behind admitting either of true and false as
return values for this case?

It seems to me that NULL (as in SQL) is most appropriate here. But
that's not boolean logic any more, so it would have complicated things.

In theory, an IllegalCrossTreeNodeComparisonException would also be
conceivable. But I wouldn't want to have to do exception handling in
XSLT because it complicates things.

A recoverable error would be similar in nature to leaving the return
value up to the implementation.

The user can always use the root() function to check if both nodes
really originate in the same document.

<xsl:stylesheet version="2.0"
  xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
  <xsl:output method="text"/>
  <xsl:variable name="t1"><Urmel/></xsl:variable>
  <xsl:variable name="t2"><Urmel/></xsl:variable>

  <xsl:template xml:space="preserve" match="/">
<xsl:value-of select="generate-id(   /*)"/>
<xsl:value-of select="generate-id($t1/*)"/>
<xsl:value-of select="generate-id($t2/*)"/>
<xsl:value-of select="$t1/*   >>     $t2/*"/>
<xsl:value-of select="$t1/* &lt;&lt; $t2/*"/>
<xsl:value-of select="root($t1/*) is root($t2/*)"/>
  </xsl:template>
</xsl:stylesheet>

--
Michael Ludwig