RE: How to specify a Processing Instruction? (better: how tocontrolencoding on saving)

[Julian Reschke <julian.reschke@gmx.de>]

> From: Arnold, Curt [mailto:Curt.Arnold@hyprotech.com]
> Sent: Wednesday, August 29, 2001 11:39 PM
> To: 'xml-dev@lists.xml.org'
> Subject: RE: How to specify a Processing Instruction? (better: how to
> controlencoding on saving)
>
>
> Julian Reschke wrote:
> > > Treating the XML Declaration as a Processing Instruction in
> > the DOM is
> > > a MSXML'ism.  Not expressly forbidden in the DOM spec, but
> >
> > I'd say it is forbidden. For instance, when loading the same
> > document into two different DOM implementations I'd expect
> > that they create the same number of processing instruction
> > nodes. Otherwise they won't be ineroperable.
>
> >From http://www.w3.org/DOM/faq#XMLdecl
>
> Currently (Level 1 and Level 2) there isn't a DOM object to
> represent either the XML declaration or the text declaration. As
> a consequence, some implementations of the DOM do not model the
> XML or text
> declaration at all, while others model them as processing
> instructions (which they resemble, although the Infoset declares
> them to not be PIs). Neither interpretation is incorrect with respect to
> Level 1 and Level 2 of the DOM.

Interesting. I don't like that interpretation, even if it's from the FAQ.
For instance it means, that XPath implementations running on top of the DOM
must add special code to filter out PI nodes that are XML declarations.
Ugly.