RE: LC-16 ( LC-132 ): Allow arbitrary order with occurrence > 1

> -----Original Message-----
> From: www-xml-schema-comments-request@w3.org

> >>Could
> >>you give a concrete use case for allowing an arbitrary sequence of
> >>a, b, c, and d elements where (a) the sequence of the elements is
> >>significant,
> >
> >Did you want to write 'insignificant'? That's what both the current
> >all groups and my proposal are about.
>
> I think not.  If the sequence of child elements has no significance,
> and they are not all optional, then the order of children might as
> well be (and usually should be) fixed.

Here is my list of use cases for all with occurs > 1:


CASE 1
======
Many specification languages have an unordered different kinds of
declarations to be interleaved in any order, but have one or more
distinguished types of declarations that can occur only once.  Usually this
happens when:

a)	The entire file represents the flattening of some heirarchy or DAG of
relationships.  The "distinguished" delcaration would typically correspond
to the root; and

b)    The XML containment heirarchy is used to express a _different_
relationship.

Consider Java source files, which should contain at most one public class,
and C source files, which can contain at most one main function.  It's a
reasonably common pattern.

And the order of the declarations often _is_ significant.  In a rough
descending order of significantness:

i)	It is reasonably common to restrict declarations to refer only to
previous (not following) declarations.  This is common, for example, when
you're flattening a DAG of relationships, and you want to enable streaming
processing.  It's also very common in programming languages.  In general,
the position of a declaration in the source file contributes to the context
in which the declaration is interpreted.

ii)	The order is not semantically significant, but affects the efficiency of
processing.  If a declaration is referred to in 100 places, for example, it
will often be more efficient to put it first, allowing the system to
remember just that first declaration, instead of the 100 referencing
declarations.

iii)  The order is not semantically significant, but affects readability.
(Rick Jelliffe's argument) This is important for long specifications.  How
come XML schema doesn't force complex type decls to come before element
decls?


CASE 2
======

I will assume that all groups with all maxOccurs <= 1 are useful in at least
some situations, simply because we include them in Schema (I was never
convinced that they were, actually -- I subscribe to your view above).  I
know that openness-by-wildcard is useful in several situations.  As far as I
can tell rationale for using each of these features is orthogonal to
rationale for using the other, so there should (statistically) be situations
in which these features are useful simultaneously.

Openness-by-wildcard requires repetition, however, and so it is currently
impossible to use it within all groups.  The way to open an all group would
be (a & b & c & whatever*), using interleave semantics, which corresponds
exactly to Martin's desires.

Received on Monday, 16 October 2000 15:08:27 UTC