I don't how much I can add to the current discussion. I can provide some 
historical perspective on why the DITA schemas are the way they are and 
the cause of Eliot's headaches w.r.t to the DITA schemas

In its current incarnation the DITA schemas only have redefine 
statements in the head schema documents that assembles other schema 
documents.  We did this way so that we could validate the XML Schemas 
using Xerces-C.  There was a bug opened against Xerces-J to address the 
issue. - https://issues.apache.org/jira/browse/XERCESJ-1219  I don't 
know if this will ever be fixed because it will probably break some 
schemas that are currently valid. 
If anyone attempts to redefine content model that have been redefined in 
the head schema and they attempt to use those with Xerces-C they will 
get schema validation errors. 

 If a user wants to assemble a schema in a different manner add/remove 
schema document, they should create a different head schema.  Doing it 
in this manner users can have different behavior based depending on the 
head schema that assembled the components. There are parent/child 
dependencies between files some the dependencies are flattened in the 
head schema.

The current model, minus Eliot's issue for the moment, is the lowest 
common denominator that will allow the DITA documents to validate 
against the schemas against the most of the major or most commonly used 
XML parsers available so far. 

 From a DITA perspective we should be able to define some markup 
(hopefully  based on the proposal that Michael Kay to the WG) we get an 
output that will no longer makes use of the redefine mechanism but still 
make sense when dealing with head schema and its possible impact when 
storing these schema in a CMS or XML Database.

Kind regards,
Eric







From ceek63@y... Wed Jan 09 20:51:24 2008
Received: from maggie.w3.org ([193.51.