Re: [dsdl-discuss] [xml-dev] Re: Fw: Co-Occurance constraint proposal

Martin Bryan said:
> Rick
>
> Interesting.

Yes. I wanted to make the case to the XSD WG that they should move in
proven directions, towards DSDL. No idea how it will be received, but
worth saying I think. It is silly for them to add bogus
demi-Schematron-isms to XSD when RELAX NG has some feature that would do
the job from inside grammars!

Background: Apparantly the XSD WG is discussing XSD 1.1, and looking at
removing cardinality constraints on <all> and on adding minimal
co0occurrence constraints such as one attribute requiring another. It
would be a victory for sanity for the XSD WG to converge with RELAX NG in
this area rather than perpetuate some harebrained NIH-ism.

So I wanted to bring the RELAX NG approach to their attention, because I
think we (DSDL-ers) need to move from the reactive "What can we do that is
better than XSD" to the supportive "How can we get XSD to converge with
DSDL." It is a win for RELAX NG if someone chooses it over XSD; but it is
also a win for RELAX NG if XSD converges. If our long-term hope is that
XSD is layered into a RELAX NG compatible substructure with type
assignment as a super structure, then it is important that XSD is revised
in ways that don't add technical barriers.

> I'm not sure how attributes would apply rather than elements
> embedded within the schema's appinfo element. It would help if you had
> examples that showed the use of all three of your extension framework
> "places". I'm also unclear what would happen if elements and attributes
> were
> associated with the same validatable element.
>
> Another point I am unclear about is how and when you would add element
> names
> to the Extended Validity Attempted attribute. Is this done as you stream
> in
> to indicate that the element needs to be validated against these types of
> rules, or after validation, to indicate what rules have been applied? What
> happens if there is a name in the Extended Validity Diagnostics that is
> not
> included in the Extended Validity Attempted list?

The post was to the XSD group. Changes to XSD have to be expressed in
terms of their system. The XML document disappears after parsing and
instead we have an infoset with non-XML properties, the PSVI. So we don't
add element names of attributes, the Extended Validity Attempted is a
property that all attribute have in the PSVI (in my proposal).

> Which brings me neatly onto my last point. The use of namespace as the
> first
> member of the Extended Validity Attempted list is confusing. It seems to
> imply that only one element of a specific type can be used in each rule
> set.
> Surely this should not be the case? What you probably want is either a
> unique identifier of the relevant rule, or a path that identifies which
> rule
> in a set it was.

The PSVI listing is exhausting but not exhaustive. I only put what I
thought was the basic information there for loosely coupled schemas: text
messages tied to a location, with an indication of the source schema
namespace for each message. I think I have a note saying that
implementations would certainly add more info. The PSVI is just a
formalism anyway, because there is no standard API for it (or, at least,
only some platform specific ones).

Cheers
Rick

--
DSDL members discussion list
To unsubscribe, please send a message with the
command  "unsubscribe" to dsdl-discuss-request@dsdl.org
(mailto:dsdl-discuss-request@dsdl.org?Subject=unsubscribe)