XMI and core components

[Chris <chris@cnelson.demon.co.uk>]

In message <02a801bfc6ed$487f14e0$20cc66c3@sgml>, Martin Bryan
<mtbryan@sgml.u-net.com> writes
>>
<SNIP>
>This is the main criticism of XMI. Without structured messages you cannot have 
>structured queries. Without structure you end up with having to adopt a naming 
>scheme that differentiates the different contexts within which a data object 
>occurs. When you name object based on their context they are no longer reusable 
>(see BSR for examples). Reusable core components need to be context 
>independent. 
>They need to be placed in contexts within messages. Therefore the messages need 
>to be hierarchical. Therefore XMI is not suitable for ebXML.
>
>Martin Bryan
</SNIP>
>
I will wait until the "smart" core components and extensible mechanisms
are defined before I take a judgement on this.  I think this design will
be an important deciding factor on how to map UML to XML, and whether
XMI has a role to play.  I believe that both XMI and "simpler"
hierarchical representations can co-exist in the overall architecture as
standard XML exchange mechanisms (depending on the requirement), just as
EDIFACT and ANSI X.12 equivalents will need to be supported by the
overall architecture.  This is what the statistical (EDI) domain
require: the investment has been in the model and the next generation of
systems that process the "messages" will do so in terms of the objects -
it is a software problem to transform the incoming syntax representation
(XML, XML/XMI, EDIFACT) to the objects and vice versa.  Developments are
already in place at the level of the community of users to put this
"architecture" in place.

Chris Nelson