[
Lists Home |
Date Index |
Thread Index
]
Paul Prescod <paul@prescod.net> wrote:
|> http://lists.xml.org/archives/xml-dev/200002/msg00609.html
|
| That's not an *out-of-line* declaration.
I assume you're talking about the two PIs at the top. They're necessary
just as declaration syntax can be necessary - to explain the markup
actually used in the document.
| It is inline and intrusive for many of the same reasons namespaces are
| intrusive.
The only "inline" markup is (possibly) the 'xdc' attribute in my example.
With an out-of-line mechanism, yes, such markup wouldn't have been needed
in the instance.
But there is another important factor here. That multiple taxonomies are
simultaneously in scope is ultimately a feature of the instance document.
It is gratuitously restrictive to require that this be defined only in
terms of "document types". One could *design* a document type and within
its definition explicitly specify the external provenance of its various
parts, but that doesn't cater to one-offs or to situations of limited
applicability - such as when you don't want to associate an element type
with another, but rather to associate only specific instantiations of the
element type with some other "template". This key feature of AFs - that
it's up to a document *as an instance* to say how it maps to any exemplar
- gets lost more often than any other.
As for means, you mentioned having a CSS-like mechanism. In the same
philosophical vein (i.e. decorating a tree) I'll mention LPDs.
But you won't get away from an irreducible minimum, which is to declare,
somehow, within the instance - that is, er, inline - that such ancillary
information is in play.
|
