[regrep] Re: Core Components Specifications

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

Message text written by John Yunker
>
As such your comment about simple XML mechanisms is right on, however I
think that we need to not only understand the limitations of both XML and
UML, but more importantly understand and leverage their power working
<<<<<

John,

Thanks for the comments.  I'm not disagreeing with this.  Of course people
that want to do OO and modelling will turn to UML.  Then the semantic
dispersion problem is one that we've been grappling for years.

UML is not a major silver bullet here either.  The point is that the
solution
is a blend of technologies and techniques.  Mandating UML and believing
this will ultimately solve all evils and semantic dispersion is fantasy 
dreaming.   For starters UML was never designed to do that job - only
part of it.

Also - as you point out UML / XMI /RDF is not ready yet.   My assessment
is that it will be another two years before that is sorted out and
functional.

So - pragmatic approach is CCR/CRI right now.  This works with the 
registry and allows people to migrate their existing components to
a coherent structure and method within the context of deriving core
components - using simple XML and web forms interfacing - and
building tranditional transaction based processes.

The industry is crying out for a simple solution they can use 
immediately - they will measure ebXML on that ability to provide
a solid set of business transactions that are compatible with 
todays deployed technologies and CCR/CRI does exactly that.

Once we have that underpinning - then two years from now you'll be
well placed to apply UML once it is fully capable - AND you will help
guide and direct that UML work so that it can work with ebXML.

If UML was wonderful, easy and capable of doing what we need 
right now I'd support it whole heartedly and everyone would be
using it already - they would not need us to be mandating they do it.

Fact is it is even more immature than ebXML, and to bet the farm on it -
and thus requiring people to invest in yet another technology is 
not sound business implementation sense.

Notice - the recent HR-XML session 
in Orlando before the main XML2001 conference - had 
implementers detailing exactly how they tried UML - and it 
failed.   They then discarded it and built a solution that 
worked based on XML only  (see the article on XML.com).

So I'm basing this on not just my own experience but
also authenticated large deployment efforts from 
people who really wanted to use UML - but could not.

We're not trying to sideline UML here - but provide a pragmatic
and real implementation today - that will also support those
who want to use UML.   

The challenge is to make these two worlds work together, 
and we are making excellent progress on that thru the 
CCR/CRI working group and the liaison work with the UML/XMI/XML.

We do not need people external to the process trying to mandate
one approach over another.  Both are needed.

DW.