I've made an attempt at implementing the WG's decision wrt <element ref=...
I have found a number of things:
<element
ref='...'
besides minOccurs
and maxOccurs
;<attribute ref='...'
, including the fact
that in one of the schemas I looked at principled use was made of
global attribute declarations which were sometimes refed simply,
resulting in an optional declaration, and sometimes refed with a
use='required'
attribute;use
on <attribute
ref='...'
is very modest, without significant ramifications.Accordingly, I request the WG to reconsider its decision on this issue, and instruct me to proceed as follows:
ref
attributes so they must not specify any of the problematic
properties (nullable
, value constraint
,
identity-constraint definitions
, disallowed
substitutions
;ref
attributes so they must not specify value constraint
;(1) above eliminates the need to have modified element declarations;
(2) and (3) eliminate the need to have modified attribute declarations while retaining the parallel between elements and attributes.
This option not only is easy to implement, it leaves no barrier to subsequent improvements. I don't believe there is any other solution with either of these properties.