uff... I can't resist. I open my legs to ridicule:

I think using XSL on XML is like sailing (Sax is like surfing) -- that is
my imagining -- at least it helps me :). You basically do what you can
given the water (standards), wind (XML) your boat (XSL)  - but you still
have to get to your destination.

 1. There is no need to have any source XML document.
     http://www.w3.org/TR/xslt20/#initiating

 2. There is no need to "get to your destination". You may be
producing results non-stop all the time and they may be used by
consumers (say in a pipeline) as soon as a new result appears. Here by
"result" it is good to mean anyone of the possibly unlimited number of
final result trees that the XSLT 2.0 Recommendation allows a
transformation to produce (for example using <xsl:result-document ...
/> instructions)
    http://www.w3.org/TR/xslt20/#element-result-document

 3. A good tail-recursive implementation can guarantee that
"unlimited-stack-depth" recursion will be performed naturally without
any problems such as "stack overflow"

 4. It is not an absolute requirement for such a transformation to
access an "infinite data structure" although if necessary such access
can be implemeted using a "on-demand", or a "partially produced" one
using a lazy evaluation approach.

--
Cheers,
Dimitre Novatchev
---------------------------------------
Truly great madness cannot be achieved without significant intelligence.
---------------------------------------
To invent, you need a good imagination and a pile of junk
-------------------------------------
You've achieved success in your field when you don't know whether what
you're doing is work or play