Melvin Chin wrote:
> Just exactly what does NVDL disrupt?
Yes, I'd agree with that.

What perhaps hasn't come through is that NVDL is just as useful in an 
all-XSD environment as it is in some mixed RELAX NG/Schematron/DTD 
environment.

The issue of how to declare and treat wildcards and of the default 
openness/closedness of schemas has long been discussed in XSD and schema 
circles. Indeed, it has occupied Roger's attention a lot at various time 
as he has been working through issues.

One of the difficulties without NVDL is that at the moment it is the 
developer of the original schema who decides how open that schema is, 
and at what points wildcards or other vocabularies can be used.   If 
their decisions are OK for a particular user, it is fine, but if some 
other user wants to adopt some profile or some particular pattern or 
something that doesn't fit in with derivation by extension (for example) 
or have some kind of interleaving, then they must hack together their 
own schema. But the developers of a vocabulary are the wrong people to 
specify its use: indeed it results in unnecessary attention to 
structures over fields.*

Many people, especially on standards groups, are extremely loath to do 
this, because they are in effect making their own dialect of someone 
else's vocabulary. Some (ODF took this route with SVG) will just use 
their own namespace with the same local names in order to avoid making 
an independent standard.

With NVDL, the decision about how to combine schemas does not depend on 
the schema modules (if that is what the developer chooses) but is 
deferred to a higher level.  The people who make the vocabulary may have 
declared it closed, but the the adopters decide whether to override 
this. And all this without impacting the original schema.

This is validating a view of the data, of course: the idea that elements 
are actually stripped out is not necessary to an implementation.
 

Cheers
Rick Jelliffe

*) 
http://www.oreillynet.com/xml/blog/2007/11/standardize_the_jellybeans_not.html