RE: re discussion topics

["Duane Nickull" <duane@xmlglobal.com>]

<terry wrote>
No, there's no reason that you could not query *the metadata in a
registry* for all the items that match a given set of criteria;
the reason for needing a unique identifier is that you need to
be able to refer to a registered item by it.
</terry wrote>

I think this is exactly what I was trying to say. My apologies for not being
clearer.  I am in favour of the verticals being able to query for associated
value types or other criteria but for an application to query for it,
pursuant to ebXML Architecture (current draft - not final yet),  you need
that unique identifier for each item.  The UID is passed along with the
document instance to the parties to a transaction.  Their applications then
check the cache to see if they recognize the data element.  If not, their
application can send an XML string to the registry and perform and RPC to
query the metadata.  If it is not found,  a manual search can be conducted
by a human via a [web based] GUI.  That query should be able to search all
criteria of a data element.  The application and the cache stages must be
restriccted to look for ID's which are unique.  A query result of 2 items
means the process is ineffective.

>This issue is a well known rat hole.  There is no reason for EBXML to
>care whether identifiers are meaningful.  There is also no reason for
>EBXML to specify anything further about them except that they be URNs.

Functionally - yes.  I agree.  However,  I personally want to see ebXML
complete ASAP and I forsee certain things impeding rapid implementation,
this issue being one IMHO.

...
>I don't understand the example; in any event there's no reason a registered
>item cannot have more than one unique identifier - and some will have, as
>they already do.
Again, I completely agree. What I have said is that "each data element must
have at least one unique identification key".  Without at least one, it will
not work.

>| SECOND ISSUE:
>| 2) Are empty elements in an XML document instance required to reference a
>| data element unique identifier so they may be retrieved from a
repository?
>
>That's not an issue for ebxml-regrep.
It was brought up to the list today so we are discussing it.  Has this been
resolved already by Oasis Reg Rep or anyone else?

>| I think yes.  IT is important to know what the node is, even when it is
>| blank.  This would allow for less chance of error at the receiving and in
a
>| document choreography.
>
>Someone already pointed out that the definition of semantics goes in
>the schema, not the instance.
I don't think you understood the issue.  John's example was predicated by
saying " please don't pick apart the xml syntax" inferring that he wanted to
express the idea only.  We are most likely going to place the semantics in
the schema / dtd / as per the published material.

Beers!

Duane Nickull