Re: unambiguity constraint

[Murata Makoto <mura034@attglobal.net>]

Kohsuke KAWAGUCHI wrote:

> 
> P.S. I think we are looking for a constraint that makes it TDLL(1),
> rather than making it unambiguous.

I have changed the name.  The new name is "restrained competition".  
In XML Schema, a tag name within a single content model can have 
at most one type.  I call this restriction "single type", and your 
proposal is basically based on "single type".

Kohsuke KAWAGUCHI wrote:

> The following constraints might work:
> 
> 1. the parent element/attribute (of key/keyref) must has a simple name.
> 2. its name may not be accepted by any other elements/attributes that
> are in the same content model.

The same constraint should be applied to ancestors recursively.  If <grammar> 
or <define> appears, recursion stop there.  Moreover, if the key/keyref 
appears as a non-immediate subordinate of <define>, the same constraint 
should be applied to all <ref> elements that reference to this <define>.

Cheers,

Makoto