Hi Rick

Thanks for your reply.

Fair comment. In my case, however, co-occurrence constraints seem to enter in at a lower level of improtance, hence I've started (some time ago now!) with the grammar-based approaches.

However, I'd certainly have a good look at Schematron if it helped to resolve the current impasse. But from Eddie's suggestion it appears that it doesn't actually help owing to the absence of the ability to test against regular expressions. Is that correct?

The mythical RNG schema

<element name="example"> <interleave> <data type="token"> <param name="pattern">[a-zA-Z0-9]+</param> </data> <element name="a"> <empty/> </element> <element name="b"> <empty/> </element> <element name="c"> <empty/> </element> </interleave> </element>

would validate my original example (now corrected!)

<example> The quick <a/> brown fox <b/> <a/> jumped <c/> over the lazy dog 5 times <c/> </example>

but would reject

<example> An <a/> example <b/> <a/> containing characters such as *&^%$#. <c/> </example>

Can Schematron (possibly + RELAX NG) do better in this case?

Cheers