Yes, it does not seem right for most schema languages.
This is because of their pass/fail or PSVI approaches. Something is valid or invalid against a schema, rather than having a processing or semantic status assigned to it.
I suppose DTDs and XSD can model "must accept" by implying an attribute value to elements they know about. Schematron can let you assign roles to assertions or patterns, typically error, warning, note, and potentially other properties.
The most commonly used standard way to do this is Microsoft's "secret weapon" for futureproofiing which is the MCE (Markup Compatability and Extensibility) spec. It is part of OOXML but used by them in many other places. There is no equivalent in ODF (I tried to propose adopting MCE but some mix of NIH and the reality of the more limited resources of the ODF developers was against it). ISO is currently rewriting the MCE to be more implementable, I believe. MS has documentation on how they use it in OOXML at
http://msdn.microsoft.com/en-us/library/gg548604%28v=office.12%29.aspxFor MCE, the approach is to do things by markup rather than schemas. Primarily chunks or information units, not individual elements. Makes sense.
Cheers
Rick
