[relax-ng] Binary XML by using RELAX NG

[Kohsuke Kawaguchi <Kohsuke.Kawaguchi@Sun.COM>]

This is just a toy idea I'm playing with.

The idea of "binary XML" has been around for a while. One kind of binary
XML can handle any well-formed XML. There is another kind of binary XML
where the format uses schema information to further compress data.

I saw these schema-guided binary XML formats for DTD and XML Schema,
then I thought RELAX NG is even better framework for schema-guided XML
as well.

The idea is this. If you have a schema 

<element name="foo">
  <optional>
    <attribute name="xyz"/>
  </optional>
  <oneOrMore>
    <element name="bar"><text/></element>
  </oneOrMore>
</element>

and an instance

<foo>
  <bar>abc</bar>
  <bar>def</bar>
</foo>

then you can reproduce the same document from a set of "instructions"
that basically describe which branch of choice you had taken. For
example, the first choice you hit is <optional>, so the first
instruction is "skip". Then the next choice is <zeroOrMore>, so the
instruction is "twice". You can record texts and values as you go.

So the complete instructions to reproduce the binary are:

1. skip  (the first <optional>)
2. twice (repeat <oneOrMore> twice.)
3. "abc" (the contents of the <text/> in the first iteration)
4. "def" (the contents of the <text/> in the second iteration)



How is this?

regards,
--
Kohsuke KAWAGUCHI
Sun Microsystems                   kohsuke.kawaguchi@sun.com