Re: [relax-ng-comment] Mixing patterns and (empty) tags

[Rick Jelliffe <ricko@topologi.com>]

From: "Daniel Prager" <danielp@deakin.edu.au>

> The particular work I'm doing has been interesting in that on a couple of
> occasions
> so far I have come upon real-life examples that strain the expressivity of the
> schema validation languages.  To what extent is completeness a design objective
> of RELAX NG?

Schema languages tend to do two things: 

  * first, they express constraints;
  * second, they impose/exploit some analytical framework (objects, grammars, patterns,
    type lattices, etc) which allows management, declarativity, manipulation by
    higher-order functions, and efficient implementations.

These two things are pretty much opposed, and RELAX NG has been very good
at trying to take the grammar framework further than anyone else has. 

The problem (which some may think WXS suffers from) is that in order to cope
with many cases within any one declarative framework, you end up tacking on
many complicated and unsatisfactory ideas or have special cases.  That is
why it may be better to simply bite the bullet and say "I will use each schema
language for what it is best for; I have better things to do than stretch 
schema languages to their limit".   If you build in the idea that you may have
a RELAX + Schematron combination from the beginning, you may find you can
get the best of both worlds: a pair of nice relatively simple schemas.  Remembering
that the simpler the expression of some constraints are, the more chance that
people will be able to implement and maintain the system.

Cheers
Rick Jelliffe