Proposal for Triage: RelaxNG for DITA Vocabulary

[Eliot Kimber <ekimber@rsicms.com>]

George Bina from SynchroSoft (OxygenXML) has developed a technique and
supporting software for using RelaxNG for defining DITA vocabulary modules.
He has made this work available through the open-source dita-ng project:
http://code.google.com/p/dita-ng/. This project provides a complete RelaxNG
expression of the DITA 1.2 vocabulary and constraint modules and document
type shells.

George presented on his work at the DITA North America conference and those
of us who saw it were very excited.

To date, the main barrier to using RelaxNG for DITA vocabulary has been that
RelaxNG did not have a built-in mechanism for defining default attribute
values. This made it unsuitable for normal DITA use where we don't want the
@class and @domains attribute explicit in our source documents.

RelaxNG now has a companion specification that provides a way to define
attribute defaults. George has implemented processing for this specification
as an extension to the standard RelaxNG Java library (and has or will
contribute that code back to the main projects). It works with any
Java-based process that can add the necessarily libraries, including the
DITA Open Toolkit.

With the ability to define default attributes through RelaxNG, it now
becomes possible to use RelaxNG with DITA documents just as you would DTDs
or XSDs.

That's useful in and of itself.

But what really got us excited was the way that RelaxNG's design makes it
about as easy as it could possibly be to do module integration and
constraint.

Here is the RelaxNG for the module inclusions for the topic document type
shell

  <!-- Include modules -->
  <include href="../../base/rng/topic.mod.rng">
    <!--
      Redefine the infotype entity to exclude
      other topic types and disallow nesting
    -->
    <define name="topic-info-types">
      <ref name="topic.element"/>
    </define>
  </include>
  <include href="../../base/rng/highlightDomain.mod.rng"/>
  <include href="../../base/rng/utilitiesDomain.mod.rng"/>
  <include href="../../base/rng/indexingDomain.mod.rng"/>
  <include href="../../base/rng/hazardstatementDomain.mod.rng"/>
  <include href="abbreviateDomain.mod.rng"/>
  <include href="uiDomain.mod.rng"/>
  <include href="programmingDomain.mod.rng"/>
  <include href="softwareDomain.mod.rng"/>


That's it. Each module declares how it contributes itself to the larger
whole like so:

The @domains contribution:

  <define name="domains-atts-value" combine="choice">
    <value>(topic abbrev-d)</value>
  </define>
  
Content model extensions:

  <!-- Define domain extension patterns -->
  <define name="abbrev-d-term">
    <ref name="abbreviated-form.element"/>
  </define>
  
  <!-- Extend the patterns with the domain contribution -->
  <define name="term" combine="choice">
    <ref name="abbrev-d-term"/>
  </define>
  

This is simply saying that the <term> element's content is extended by
adding <abbreviated-form> as an additional choice.

What this means in practice is that the integration of modules is done
simply by referencing the modules you want--each module manages its own
integration details. There is no separate integration configuration
activity.

This makes RelaxNG very attractive as the primary form of vocabulary
definition *as long as* one can then reliably generate the DTD and XSD
versions of the vocabulary. George is working on implementing that process
now and I will probably do some work on it as well.

Personally, I'm very excited about this because it gives me the potential to
greatly simplify the task of defining and maintaining large vocabulary sets,
such as DITA for Publishers and allows the easy provisioning of both DTD and
XSD forms.

RelaxNG also provides for embedding Schematron rules directly in the RelaxNG
documents, meaning that a single document can define both the base XML rules
and additional rules checkable through Schematron processing (this is also
possible in XSD 1.1 with its new "assertion" feature, but support for XSD
1.1 is still not universal).

RelaxNG also provides for embedded documentation (similar to how you can do
it in XSD). 

Even if you never used the RelaxNG modules directly with documents, they
would have tremendous value as the development source for the working
modules.

Given all that, I would like to propose the inclusion of RelaxNG as a
recognized DITA schema language in addition to DTD and XSD. I am more than
willing to act as champion for this proposal and do the editorial work.

Cheers,

Eliot


-- 
Eliot Kimber
Senior Solutions Architect, RSI Content Solutions
"Bringing Strategy, Content, and Technology Together"
Main: 512.554.9368
www.rsicms.com
www.rsuitecms.com
Book: DITA For Practitioners, from XML Press,
http://xmlpress.net/publications/dita/practitioners-1/