[
Lists Home |
Date Index |
Thread Index
]
> > > The normalization of CR, LF and CRLF can be avoided by using their
> > > character entities. So I still can preserve fidelity.
> >
> > I don't think this is true. For instance, using MSXSL, try:
>
> I don't think using an implementation of XSLT is a valid way to
> say what the XML spec says should be done. Its always dangerous
> to use an implementation (with possible bugs) as a means of saying
> what the spec says.
Let's be clear. The XML specification describes (in effect) how an infoset
is constructed from a source character stream, and it explains that native
CR, CRLF, etc are normalized to x0a, while character entities such as 
are not normalized. The XSLT specification describes how an infoset is
serialized to a result character stream, and the only thing it says about
how x0a and x0d characters should be serialized is the statement that the
serialized XML must parse back to the same infoset. This means that if the
result tree contains an x0a character, it can be serialized interchangeably
as CR, CRLF, LF, or 
 but if it contains an x0d character, it must be
serialized as 
.
So my reading is that the character reference 
 will not necessarily be
preserved through an XSLT transformation (it may end up as CRLF), while the
character reference 
 should be preserved.
I'd be surprised if Saxon gets this right...!
Mike Kay
|
