[dsdl-discuss] Re: Potential use cases for complex value validation from ebXML

From: "Martin Bryan" <mtbryan@sgml.u-net.com>

> Agreed, but the problem is that the constraint we want has to be expressed
> at present as "use this subset of that enumerated list maintained by the ISO
> 3166 Maintenance Authority in their HTML/PDF files". Until we get them to
> maintain DSDL conformant enumeration lists we can't start to do anything :-(

(If they maintained their lists in an XML format, then a Schematron schema
could point to them.)

This problem also arose for XML 1.1: they wanted to to defer to Unicode but
not have to track Unicode changes explicitly. So they (initially) decided
to loosen the naming rules in order to give leeway to Unicode to change.
So they give up the ability to do specific checks, which is not really
the best idea for a schema language!

We have to be clear what we are providing by a reference to a controlled
vocabulary: is it the current list or the list at the time the document was
made?

So I think simple referencing external specs is a wrong-headed approach:
because it fails to come to grips with a central validation semantic. W3C can
get away from this issue a little, because they are more concerned with
temporary documents zipping around or with RDF statements, rather than
markup of documents that may have very long life.

I think what we need to do for every non-SC 34 controlled vocabulary that
we use is to:

1) Provide in an annex a normative "reference implementation" or "XML
binding" of that controlled vocabulary, comprising all the codes that have ever been
used. Then maintain a non-normative public version in which codes are
added but never removed. Call them a namespace or something to distinguish
it from the standard. Implementations of DSDL use this list as their default:
it means that any correct document that uses a code will pass, but it does not
check that all codes are from the same period. Specify that implementations
may augment this list (e.g. merge the one from the web) if new codes come
out and the reference list has not been maintained.

2) Provide a mechanism by which implementations are allowed
as an option to use a particular version of a controlled vocabulary,
but this is by their own provision, not by ours.

3) Try to get the maintenance bodies to provide this list themselves.

Controlled vocabularies that are updated or obsoleted pose
difficulties for documents: a document that was valid at a
certain point in time should not become invalid just by the
passage of time *unless* the validation specifically includes
the semantic that the controlled vocabularies are the
most recent ones only.

Cheers
Rick Jelliffe

--
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)