> Until now I was under the impression that the "instance of" operator
> is usable only if the value has been type-annotated as result of
> validation against a specific xml schema (as described on page 287 of
> the XPath 2.0 book), (which may not be the case with a significant
> part of all transformations written and used in the real world).
> 
> Is my understanding of the applicability of the "instance of" 
> operator wrong ?

It's incomplete... For atomic values, you can simply do things like

123 instance of xs:boolean  => false
"123" instance of xs:string => true

Also for nodes, you can do

$node instance of comment()
$node instance of element(para)

neither of which requires a schema.

Michael Kay
http://www.saxonica.com/