> > If I shall reserve any character for a purpose of character map then I
> > should replace those characters in those literals.
> 
> but if you use private use area characters or unallocated unicode 
> characters for this then they should not be apperarig in your surece 
> documents anyway. (What would tehy mean?)

I realize that I might be too pessimistic in this regard, but if it's a 
valid unicode code point (even if it's reserved) then it could appear in 
the input for the generation.

> Or, you could use an xml 1.1 parser and then use #1 #2 #3 for this, then
> they may not appear as characters only as character references so again 
> should not be in your source documents.

That's tricky. 
When one assumes xml 1.1, one should asume that input is also xml 1.1, thus 
permits those characters.
--
Vladimir Nesterovsky
http://www.nesterovsky-bros.com/