Re: [relax-ng] Limitation in the compact syntax

[James Clark <jjc@jclark.com>]

> While trying to translate Makoto's ooRelaxNG examples into the
> compact syntax, I bumped up against a problem: once inside an
> annotation, you can only use annotation syntax.  This doesn't work
> well for things like ooRelaxNG, where the os:class element can contain
> snippets of RELAX NG.

ooRelaxNG isn't RELAX NG.  It is a distinct vocabulary, having a lot of 
similarity to RELAX NG, which can be easily translated into RELAX NG.  The 
os:class element is not a RELAX NG annotation, rather it's a first-class 
member of the ooRelaxNG vocabulary, just like os:element (note that it's 
os:element not rng:element).  This seems out of scope to me: the goal of 
the compact syntax is to be a syntax for RELAX NG.

If you want a nice compact syntax for ooRelaxNG, then I think the way to do 
it is to take an analagous approach to the approach Makoto took with the 
XML syntax.  Design an extended language in which OO constructs are 
first-class syntactic constructs, not annotations.  Makoto's example might 
look something like this:

abstract class animal { }

abstract class mammal extends animal { }

class cat extends mammal {
  attribute color { text }?
}

class fortuneCat extends cat {
  attribute raisedPaw {
    ## inviting money
    "left" |
    ## inviting people
    "right"
  }
}

start =
  element schemaLanguage {
    element mascot { class animal },
    element specifications {
      list { "string"* }
    }
  }
}

Yes, this is a lot more work than trying to fake it up with the compact 
syntax and annotations.  This is one area where XML wins: XML tools make 
this a lot easier to do in an XML syntax than in a non-XML syntax.

James