> > When you need to apply regex matching to text that crosses node 
> > boundaries, in the past two approaches have been proposed:
> >
> > (a) create a string in which the node boundaries are represented by 
> > some recognizable textual markup (you could use saxon:serialize()), 
> > then apply the regex processing, then reinstate the node structure 
> > (e.g. by using saxon:parse()).
> 
> Provided the <note> elements don't break a sentence then it's 
> not needed is it?

No, it isn't, hence the "When" in my first paragraph... I didn't know from
the description whether this applied to the case in hand.

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