xsi:nil doesn't cancel out which attributes are allowed, so the typing is
still relevant.

Additionally, I think it's relevant to consider the philosophy of the
working group.  I've never been on this working group, so I certainly don't
speak for them.  It seems to me that they see xsi:type and schema typing in
general as providing typing information that might be useful in various
tools such as editors for displaying types differently, or extracting parts
of the document based on type, or even conversion to object models such as
JAX-RPC, or combined with specs such as XSLT 2.0.  Since the typing system
isn't just for validation, it makes sense to me that abstract types can't
be instantiated.  In a purely validation-oriented way, I agree that this
seems silly, except for my first comment about attributes.


Dennis Sosnoski <dms@s...>@w3.org on 08/29/2007 12:55:38 AM

Sent by:    xmlschema-dev-request@w....


To:    xmlschema-dev@w...
cc:
Subject:    Abstract element and xsi:nil



Is it possible to define an abstract element with nillable='true', then
using the element name in an instance document with xsi:nil='true'? This
would appear to be prohibited, but leads to the slightly absurd case
where you have to pick one of the non-abstract members of the
substitution group to say that the value is really not present in the
instance document.

  - Dennis




From S.Yona@F... Wed Aug 29 12:14:12 2007
Received: from lisa.w3.org ([128.30.52