> > It'd be much better to have that normalization be done by layers
> > above XML than to hard-wire such requirements into the core
> > of XML. I'd certainly expect normalization rules to evolve.
>
> Actually, it might be better to do it below the parser... closer to the input
> stream level.
That'd mean "inside the parser", because of autodetection
of charsets. Unless there's a standard non-text I/O layer
that XML parsers must start to use ... :)
But the actual encoding is visible through the infoset, so this
normalization (and its evolution) can be layered.
- Dave
