I note that XML 1.1 requires XML parsers to recognise all of
LF
CR
CR LF
NEL    (the C1 control meaning "next line")
CR NEL
LS     (the Unicode character "line separator")
CR LS
as equivalent (with LF as the canonical form).
My own XML parsers can't do this without a fairly major rewrite,
because they assume UTF-8 input and provide UTF-8 out without
ever expanding it, but you have to expand the UTF-8 to recognise
NEL and LS.