A really micro schema language

Here's an idea I was playing around with a while ago.  It relates to the
PossibleChildren property John mentioned.

Imagine a really, really simple schema language that

- uses a non-XML syntax;

- allows a parser to do reasonable error recovery (and thus markup
minimization, if you ignore the errors);

- allows editors to provide a useful-level of authoring assistance.

I think the starting point would be being able to say which elements are
possible children of other elements.

What might this look like? Here are some examples.

  ul / li

This is schema that says that a ul element can have any number of li
elements.

A ul element may have any number of li child elements.

  ul, ol / li

This is equivalent to:

  ul / li
  ol / li

The ul element and the ol element may each have any number of li child
elements.

  p / b, i

This is equivalent to:

  p / b
  p / i

The p element may have any number of b child elements and any number of i
child elements.

  p / .text

The p element may (directly) contain text. An element that can
contain child elements is automatically allowed to contain whitespace
between
those elements, so this really means that the p element may
contain non-whitespace text.

An img element must have a src attribute whose value has the url datatype.

  /html

An html element may be the root element.

  math / *

A math element may have child elements with any name.

 p !/ p

A p element must not have a p child element.

 br !/ *

A br element must not have any child elements.

 br !/ *, .text

A br element must not have any child elements and must not have any text
content.

I think you would need to handle attributes as well, which can be done
without too much complexity.

  a @ href

The a element may have an href attribute.

  a @ href, type

This is equivalent to:

  a @ href
  a @ type

A next step might be to use // to say things about descendants.

 form !// form

A form element must not have any form descendant elements.

Obviously, this is far from fully baked.

James