Michael Kay wrote:
> Only, I think, by forcing the "arbitrary model" to implement some kind of
> standard interface like SAX or DOM - and that would defeat the whole
> purpose, even if it were an improved SAX or an improved DOM.
I think such a design could be quite a bit simpler than DOM since it could:
1. Be read-only
2. Only support what's necessary for the XPath data model
This would be a lot easier for other object models than implementing
full DOM. Whole classes would be cut out: DocType, CDATASection, Entity,
EntityReference, etc.; and we could probably eliminate more than half of
the methods in the classes that remained. Not only would all the mutator
methods disappear. We could get rid of all the non-namespace aware
methods. Finally, since this would be intended for machine use rather
than for direct programming, we could eliminate a lot of convenience
methods as well.
--
Elliotte Rusty Harold
XML in a Nutshell 3rd Edition Just Published!
http://www.cafeconleche.org/books/xian3/
http://www.amazon.com/exec/obidos/ISBN=0596007647/cafeaulaitA/ref=nosim
