>(3) recursion over the items being summed,
>.... None of these can conveniently be used in an XPath
> (as distinct from XSLT) context.

I recently had to do something similar to this for a problem defined as:

every line of text up to and including X lines of text must have no
more than Y characters in the   line.

(my recollection of the problem is perhaps a little fuzzy, it was
several months ago)

For example something like the following:

<sch:assert test="string-length(substring-before(concat(.,'&#010;'),'&#010;'))&lt;5">1:
Every line of the <sch:name/> text node should be no more than 5
characters long</sch:assert>
<sch:assert test="string-length(substring-before(substring-after(concat(.,'&#010;'),'&#010;'),'&#010;'))&lt;5">2:
Every line of the <sch:name/> text node should be no more than 5
characters long</sch:assert>
<sch:assert test="string-length(substring-before(substring-after(substring-after(concat(.,'&#010;'),'&#010;'),'&#010;'),'&#010;'))
&lt; 5">3: Every line of the <sch:name/> text node should be no more
than 5 characters long</sch:assert>

and so forth

The assertions are only called in cases where the string length is incorrect.

As such the rather impossible to write or maintain assertions can be
generate without need to loop or recurse over an input.

I guess I would call it a Stupid Schematron trick, basically in a
language that does not allow recursion or looping such as Schematron
with XPath 1.0 looping can often be handled as part of a code
generation process when the actual constraints of the input data is
known.

Not sure if that is a good thing.

Cheers,
Bryan