Interpretation issue regarding PI nodes

I'm uncertain how PI nodes which are not included in the input node
set should be processed. As far as I understand the specification,
given the document

  <doc><a>Hey</a> <b>Ho</b></doc>

and a node set containing only the <a> element, the output of
canonicalization will be

  <a>Hey</a>

which is fine.


However, given the document

  <doc><a>Hey <?a pi?></a> <b>Ho</b></doc>

and the same node set (still only the <a> node), what is the output?

Is it

  <a>Hey </a>

or

  <a>Hey <?a pi?></a>


Whether this merits an erratum I don't know, but I would certainly
appreciate knowing the answer. :)

-- 
Lars Marius Garshol, Ontopian         <URL: http://www.ontopia.net >
GSM: +47 98 21 55 50                  <URL: http://www.garshol.priv.no >