Schemas and the open world assumption.

Hello everyone.

Sorry to be 5 years late, but it is just today that I question myself
about derivation by extension and closed world assumption.

Please do not hesitate to comment the following points:

If I define "a" as being a sequence of b,c:
<a>
 <b/>
 <c/>
</a>

and i extend "a" into "aa" that extends that sequence with d:
<aa>
 <b/>
 <c/>
 <d/>
</aa>

then any "aa" will not validate against the definition of "a".
right?
this sounds like a MAJOR difference with OO paradigm (where any "aa" is
also a "a").

that is what i call the closed world assumption in xml validation.

considering i need a more open world approach, i plan
to relax my schema by defining "a" in this way:
<a>
 <b/>
 <c/>
 <xsd:any>
</a>

then i feel like i could extend my "a" definition  without breaking
the "subclass" philosophy.

can anyone comment that point of view?
i am especially interested in possible pitfalls i could have missed in
using the "any" statement.
i am also interested in best practices when defining modular expandable models.

any help is very welcome.