Re: XML Syntax/Semantics

[Christopher Ferris <chris.ferris@east.sun.com>]

Typically, this is handled through the employment of metadata and models.
It is usually not necessary to embed the "description" of an element (or even
a reference to it) in a message. Rather, the message is modelled and the model
(and its associated metadata) is registered and stored in a repository.

Namespaces can be used to identify the model which defines a given message
type and used to retrieve the model and associated metadata which might include

semantic x-reference information.

IMHO, our role should be to provide the content for the semantic meaning of the

elements we include in the header, etc. as well as any information related to
semantic equivalence with other existing standards such as RosettaNet, EDI/INT,

etc. Core and/or Registry/Repository groups should define the model(s) and
meta-model(s)
for what goes in the registry/repository as the same model(s) can be employed
for
the vertical message body definitions.

XPath, XLink and XPointer will likely play a role in how the metadata might be
associated with a given element in a specific message, but I rather think that
we need not worry over this other than to monitor the work of these other
groups
and/or provide input into the development of their requirements such that this
issue be resolvable in the context of the overall architecture.

It might be worthwhile for our efforts to use a modelling tool such as Rose to
capture the metadata and establish the various semantic relationships, etc.
but we could also do this after the fact, once the Core and/or Repository WGs
have defined their models.

My $0.02,

Chris

"Miller, Robert (GEIS)" wrote:

> MILR: Responses imbedded.
>
> Bob,
>
> I think there are a couple of alternative solutions to these issues:
>
> MILR: Just a couple?!
>
> 1. XPath
>
> MILR: A 'Semantic ID' that used XLINK/XPOINTER/XPATH syntax to access
> semantic information as needed would be deep cool, and sounds very XML
> smart.
>
> 2. namespaces
>
> MILR: Namespace can help avoid element name clashes, but IMO it doesn't
> addresss the equivalence of semantic entities across namespaces.
>
> Each option should be explored for possible appication within the ebXML
> header specification.
>
> I firmly believe the issues you raise, relative to the actual business
> transactions, needs to be addressed by the parties responsible for
> specifying these standards (e.g. OTA, et al).
>
> MILR:  I believe ebXML must adopt some XML syntax rules that adequately
> address the issue of semantics.  If there is a best practice already
> invented that addresses this, we could adopt it.  I've seen several examples
> of concrete XML syntax in the efforts of others, but I've not seen one that
> adequately addresses the issue of semantics.  The X12 TR acknoweldged the
> need, and took a first stab at a syntax. I'm convinced a better solution is
> needed and feasible.  IMO, we in ebXML have work to do on this topic.  If
> one examines the diagram developed by the Technical Architecture group, it
> would seem the XML stuff falls in the 'Syntax/Schema' box in the 'Common
> Technical Architecture'.  If folks outside the current ebXML community have
> an interest, then by all means they should participate in the effort.
> Within ebXML, I believe the work could be done either within one of the
> existing teams, or in some new team.  I believe the work effort would be
> hindered if split across teams.
> I would like to think that the syntax/semantic solution would be such that
> it could be retrofitted on existing XML/EDI DTD's, such as those of
> RosettaNet, BizTalk, et al.  Just identify the semantic units,
> register/reposit (how's that for a word?) new semantic unists as necessary,
> and lay into the existing DTD/Schema the Semantic ID pointers.
>
>
>
> Regards,
>
> Dick Brooks