Re: [xml-dev] Just nodes please

[Michael Kay <mike@saxonica.com>]

> Such a rule would make it impossible to have the behaviour common in 
> many document formats (notably xhtml) that by default child elements 
> are processed but unknown attributes are ignored.

Many formats allow open content for elements as well as attributes. 
That's one thing namespaces are good for: process content in namespaces 
you know about, ignore content in alien namespaces.

> That model is pervasive throughout XML, for instance in the 
> string-value of XPath, the string value of your first example is "" 
> the string value of your second is "\nnode3\n" so these can never be 
> equal without completely rewriting XPath. 

The XPath string value concept is designed for mixed content, and it 
really doesn't work well for element-only content (which is why XPath 
2.0 makes it an error to get the string value if the element has been 
validated with element-only content). Perhaps what we really need is for 
mixed content to be properly distinguished from structured content, as a 
replacement for the current child-vs-attribute distinction.

Perhaps we should recognize that JSON does the structured content quite 
nicely, and we need to blend in XML's ability to do mixed content. 
Perhaps a format that extends JSON by putting XML-style markup in 
strings, and extends XML markup by putting JSON in attribute values? At 
the same time converging the JSON "string":value notation with XML's 
name="value".

Michael Kay
Saxonica