Serialization of XDM

[Pavel Minaev]

On Mon, Sep 14, 2009 at 1:06 PM, David A. Lee <http://x-query.com/mailman/listinfo/talk> wrote:
> Somewhat Ironically, this solves the same problem the same way JSON did i.e.
> make the serialization format the same as the Code (in Jason's case
> Javascript).
> It also has the same problem though.
> It requires an XQuery paser to parse.  Suppose you want to send the results
> to something that isnt XQuery, say as an external variable to an XSLT or
> XPATH processor.   Those would require XQuery to pase it.

Just as with JSON, the ability to parse data by eval'ing it in XQuery
is just a side bonus. Of course you'd also want specialized parsers
for this when dealing with such serialized data in anything other than
XQuery (and you'd want it even in XQuery for security reasons). But
there's no shortage of JSON parsers of all kinds, written in different
languages and for different platforms, these days. This could be
similar.

A more major advantage of using XQuery subset as a basis for such a
serialization format is that you don't need to create a specification
from scratch. For the most part, you can directly re-use XQuery BNF
and semantics from the spec.

Also, so long as it is a strict subset, any new dedicated parser could
reuse existing XQuery lexer/parser code with no changes. You'd pretty
only need a new syntax tree rewriter that would output XDM nodes, in
whatever in-memory format a particular implementation wants.

By the way, much as DOM is often maligned, the idea of defining some
common, strictly specified interfaces (in API sense) still has merit,
and having something along those lines for XDM would be nice as well
(and would mesh up well with a common serialization format).