[
Lists Home |
Date Index |
Thread Index
]
> Better to have a standard syntax and different, non-standard, independent
> object models optimized for their particular domains and uses that all
> operate on the same underlying syntax. Object models are *local*. Syntax
> is *global*.
The success of XSLT, XPath, and the (probable) success of at least the hard
core of XQuery is a strong counter-example to this assertion. None of
those operate on syntax. We can argue about whether the XPath data model
is an "object model" or not, but it ain't syntax. An XSLT stylesheet that
had to consider all the multiple possible syntactical variations on legal
input would be a nightmare. An XQuery join across an RDBMS and an XML data
store is simply inconceiveable at the syntax level.
Likewise DOM is an ugly beast, the proverbial camel designed by a
committee, but it has proved quite valuable out in the wastelands of the
Web where well-formed syntax is rare. What unifies N separate bowls of tag
soup and legal XHTML that look the same in a browser, or similar XML files
with different frostings of syntax sugar? They probably have more or less
the same representation in the DOM Core.
The XML syntax and data model are joined at the hip. (OK, there are
multiple data models, but the different flavors vary mainly with respect to
the namespace and schema type stuff that has caused no end of controversy
all across the XML spectrum). Without the data model(s) XML is useful
mainly by hard core developers skilled in regexes and formal grammars;
without the stamdard syntax those abstract data models can't be shared
without some sort of private contract about serialization. Why do we get
tangled in this permathread every month or two?
|
