|John McClure||Jul 18, 2004 4:21 pm|
|Subject:||(h) Schema languages|
|From:||John McClure (jmcc...@hypergrove.com)|
|Date:||Jul 18, 2004 4:21:49 pm|
If you'd like to copy/paste your sections (g) and (h) from your "long" memo (or something based on them) to the list, i will respond.
Section attached, as you requested. John
(h) Schema languages
Now I will address the issue of document validation because, for schema- driven applications, it determines what elements are displayed that may then be selected to be entered at any point in an element structure by the application. This is a concern primarily of those directly perfomring text entry to a schema-based application, because most other applications that might be used by attorneys are an application "system" that little doubt has a stylesheet (one consuming, i.e., reading, our XML schema) to validate the stream prior to storage of the document in archives. [I don't understand why a Schematron stylesheet' is any better than an XSLT stylesheet for this, and could argue that it's a whole lot worse.]
So Jason's concern is that an author could create a stream with more than one footer, for instance. That is, if our standard identifies such an element as a footer element by using the @class attribute, then DTD validation facilities cannot be used to ensure that there is only one footer for a <div>, for instance. XML Schema validation facilities can't be used either, because XMLS' content models also cannot be qualified by given attribute values had by the element being validated. [There is one option though that uses XML Schema's @xsi:type attribute, not the @class attribute.]
It's said that one alternative is to create a RELAX-NG schema for the standardized elements, which does allow such qualification for its content models. Certainly I think that any RELAX-NG schema should be co-publiched with standard DTD and XML Schema schemas, while shortcomings of each are discussed within the specification.
That said, their reasoning becomes this: because RELAX-NG (nor xsi-complete XML Schema) tools aren't widely implemented, it is thus necessary that we establish element names, not element attributes, to represent the concepts called for in our requirements document. Several problems with this though: (a) assumes a particular set of tools that I don't believe are typically being used to enter language of the contract (b) assumes a particular level of capability of those tools that I don't believe will persist very long (c) assumes that the tools will never provide the capability to call a stylesheet to edit text input dynamically, or to create the "lists" of options that a user has at a particular point in the document's markup. In short, it's truly poor idea to assume that APPLICATIONS will not change over the next few years to accommodate whatever the standard happens to be. In other words, it often sounds as if the discussion of what is possible -- what is desirable technically -- is beholden to how certain applications are currently designed to operate.
Of final note on this topic, is the pressing issue of getting the W3C to establish a conformance level that represents a 'middle ground' between its now two levels of XHTML 1.1 conformance, one that reflects that the operational elements we've included from XHTML 2.0 are but a constrained subset of those defined in their modules (e.g., no PCTEXT in a <section> element). They are developing I understand an XHTML 2.0 schema in all 3 of the popular dialects, each described with relevant caveats. Clearly all this relates to our ability to provide the schema-driven markup editors a schema that reflects our standard.