Most of that is spilt milk now. OTOH, PIs are open
to those who want to develop alternatives to specs
and still use XML. It is a very neat feature from
that viewpoint.
PIs were heavily criticized in days of olde as
a way to pollute content. In effect, they were
a way to avoid precisely that. But those who
used them were sidestepping the idea that all
systems should interoperate based on a formal
public declarations. That is the One World System
viewpoint. It is seductive somewhat like a
one world religion or government but inimical
to the very idea it originates in: the ability
to create information free of local system
constraints. Ultimately, the information owner
has to choose. XML, the W3C, etc., should only
have a limited ability to control those choices
and then only insofar as a system definition
enables interoperation without unduly limiting
choice. The W3C and XML are there to enable potential;
not control markets.
The owner has to choose wisely. As noted often
here, XML is as underpowered as it is because
a syntax spec enables interoperability right
to the edge of limiting other choices. Otherwise,
one becomes a wraith in the tunnels muttering
"My precious!" after surrendering all options
to the FrameworkFromMordor.
len orc
-----Original Message-----
From: Jason Diamond [mailto:]
> PIs were a common means in SGML systems to create links.
> IDEAS/IADS worked like that originally. It was the
> interpretation of some that links are not really content,
> they are a process/function. It is a theoretical issue at the
> very heart of darkness for hypertext systems.
To me, it seems obvious that the xml-stylesheet PI is not content since it's
only allowed in the prolog of an XML document (before the document element).
The xsi:schemaLocation attribute, on the other hand, is content (although it
will most likely be ignored by processors other than schema validators).
Unfortunately, just like namespace declarations, xsi:schemaLocation can
appear on any element in an instance document--though it must appear before
an element or attribute is encountered in the namespace described by the
schema found at that location. Personally, I agree with Michael and would
have rather seen something like an xml-schema PI (or a more generic xml-link
PI) in the prolog--one for each namespace the document contained. This keeps
the processor-specific data out of the document and makes it really easy to
see exactly what the document contains from a really high-level point of
view.
<?xml-schema namespace='http://example.org/'
href='http://example.org/schema.xsd' type='application/xsd+xml'?>
