[xml] Performance of tokenization?

[Bjorn Reese <breese mail1 stofanet dk>]

Does anybody have any performance measurements on how much of
the parsing time libxml spends scanning and tokenizing the text?

I can measure it myself, but I would like to avoid it if anybody
has such numbers handy.

I do not need any exact numbers, just a rough estimate (it will
end up as a single number on a slide.)