This is a recap of an possible erratum discussed within the Schema IG. 

Because attribute wildcards have a "lazy" behavior, it's possible for an 
apparently-valid restriction to violate the general notion of restriction. 
By "lazy" I mean the property that an attribute-wildcard will only match an 
attribute-information-item (AII) if no other attribute-use matches the AII. 

Best demonstrated by example: 

<xs:schema xmlns:xs="http://www.w3.org/2001/XMLSchema">

   <xs:complexType name="B">
     <xs:attribute name="FOO" type="xs:int"/>
     <xs:anyAttribute processContents="skip"/>
   </xs:complexType>

   <xs:complexType name="A">
     <xs:complexContent>
       <xs:restriction base="B">
         <xs:attribute name="FOO" type="xs:int" use="prohibited" />
         <xs:anyAttribute processContents="skip" />
       </xs:restriction>
     </xs:complexContent>
   </xs:complexType>

 </xs:schema>

I believe this is a valid restriction per the rules of restriction, but not by 
the general definition in Section 2, which includes 

Members of a type, A, whose definition is a restriction of the definition of 
another type, B, are always members of type B as well. 

B doesn't accept an element with attribute FOO="abc", but A does accept it. 
That's because A doesn't contain an attribute-use for FOO, so the attribute-
wildcard validates FOO in A. 

See:
http://lists.w3.org/Archives/Public/www-xml-schema-comments/2003JanMar/0007.html