[relax-ng] Fw: RELAX NG TC meeting 2002-10-24

[David RR Webber - XMLGlobal <Gnosis_@compuserve.com>]

Folks,

The issue that I see is simple - what is the killer application 
for RELAX NG?

I've been saying for a while that what is needed is a 
gnarly problem space that XSD is failing miserably
to solve.

The answer is there already -> eBusiness transactions.

Your reaction is "What?".  But look at the facts - the
W3C is telling people - just create an XSD and you
have instant interoperable eBusiness.  Nothing can
be further from the truth.

Unfortunately initially it does appear to people that
by writing an XSD that they are achieving a way to
exchange semantics about an interchange.  So at
first they are encouraged.   But this rapidly vaporizes
when they do more than a simple pathfinder involving
a few companies.

Then they see that XSD gives them no more, and
in some cases less, than using EDI.  And to 
compound this - at least EDI enforced a single
approach - now with XML you have 1,000+ different
ways of doing the same thing.   And worse, theres
a stack of 'band-aids' proffered to 'solve' the
shortfalls - XSLT, Namespaces, RDF, and now
the end user is in major panic mode, and doubts
over interoperablity across implementations are
a major concern.  Small wonder
that, as Josh points out, DTDs are still numerically
dominant.

The answer that ebXML has inserted to solve
this is assembly.   Assembly provides rigorous
ability to:

1. Define the explicit information structure instance
    based on business context rules.

2. Works with legacy EDI, text and other structures,
     not just XML.

3. Crosswalk between element vocabularies - allowing 
     alignment across industries / implementations.

4. Support use of central registries and web services.

5. Validation - with both pre-checks and post-checks
     captured.

6. Allows defining of catalogues of pre-built structures
    such as "Address", "Payment_Detail",  and so on, 
    (ebXML calls these BIEs - business information entities)
    that can be included simply - without the nightmare of
    namespaces - and other complexity.

7. Provide a clean and clear versioning mechanism for
    all content at element and attribute level, and code values.

8. Supports extended use of codelist mechanisms - vital 
    for EDI buy-in and support.

"So what?" I hear  us saying?!?.  How does this 
tie-in to RELAX?   

It occurs to me that if RELAX can be assembly 
pluggable - in the same way that RELAX is 
datatype library pluggable - then
now you have the ability to provide the ebXML with an
assembly implementation tied to an already mature
structure definition technology (RELAX).

The fusion of two bodies of work - providing a clear
business validation and reasons why a community
should start using RELAX ahead of XSD.

This would require an assembly parser adjunct that 
then returns the combined result of the RELAX definition
and assembly as the single instance criteria.

Note: the OASIS Registry team is providing a 
content_validation_service call in version 3.0 of
their spec's.  They envision this being implemented
as a web service call within a registry sub-system that
then invokes either an assembly template script, or
a XSLT (can handle some limited use cases) to 
return the validation results.   If you want buy-in now
would be a good chance to step-up and work with
that team on providing validation support - and
get yourselves an implementation use case and
deployed system.

OK - so I've been flogging this horse for a while.  Any
takers?

DW.
================================================
Message text written by James Clark
>> 1. RELAX NG adoption
> 
> James: XML schema is now supported by gazillions of 
> applications, MS Office apps will support it in the next 
> release; it's a complicated story but two schema languages is 
> too many
> Makoto: and MS offers commodity solutions
> James: We have no big company backing RELAX NG
> Kohuske: Perhaps we should submit RELAX NG to W3C?
> Makoto: I have heard that XHTML 2.0 will offer RELAX NG 
> schemas.
> Norm: I think it is too early to tell, many consider still 
> using both languages
> Makoto: I know that IBM has suffered from using XML Schema a 
> lot and are considering using RELAX NG
> Josh: I know that people what type derivation and single 
> inheritance, but also unordered elements
> James: They are trying to force Java on XML
> Josh: I think we should push RELAX NG to XML schema 
> translation
> Makoto: We need editor support
> James: The stability of our specs should influence future work
> Makoto: What level of XML Schema adoption is there?
> James: I don't know
> John: The DTD is numerically dominant
> Josh: People are using DTDs overwhelmingly
> James: include weak text based 
> ACTION: Write up solution (?)
<