Re: Remove all Documents? was Re: issue resolution updates

[Karsten Riemer <Karsten.Riemer@east.sun.com>]

Jamie,
we have a major disconnect.
I am not proposing any model change.
I was just responding to a number of issues that Paul had raised, calling for
additional text in the BPSS document to describe how BusinessDocuments are
composed from Business Information Objects and/or Domain or Core Components
(Don't know if I have the nomenclature right).  See issues 79, 80, 87, 92, 94,
102 and maybe others.  Since the specification schema does not provide a
metamodel for BusinessDocument structure, I don't think it is its
responsibility to describe BusinessDocument structure in text either. Paul
agreed with this in today's meeting and said that he had tried to get this
included in Core Component documents where it belongs. I do not intend to
include it in the specificatin schema. So again, this is not about modeling
changes, and is not related to any previous model related proposals.

-karsten


>Karsten, this sounds like pretty radical surgery, and at first blush I am
>having trouble conceptualizing the degree to which logical model validation
>would be achievable after that triage.   
>
>Doesn't this really get you to the same place as your e-mail from Africa?
>(Cardinality change, where transactions take 0..* documents instead of 1,2? )
>
>
>In a hypothetical zero-documents world:
>   1.   I think we would be allowing each schema (lower case "s", e.g.,
>X12, RNIF, OAG) to bring over its own level of reliability of contract
>production -- instead of imposing the constraints in the somewhat RNIF-like
>UMM.   Is that a good thing?  Certainly it would make for easier adoption
>by all kinds of other communities, including non-UMM-users of UML.  On the
>other hand, we lose some of the opportunity to logically compare apples to
>apples, and we lose the advantage of reliability and stability perceived by
>some in the binary pairing model.  
>  2.  Don't we lose some of the signals too?  I'm not sure how the DSIG
>parameter or 'receiptAcknowledgement' would work, if there is a complex or
>indeterminate association between the doc that asks for the parameter, and
>the document(s) that [respond] to it and therefore should conform to the
>parameter.    It might be a solvable problem, but it seems a little late to
>crack that tough nut in a reliable and stable manner for 1.0.
> 
>Jamie
>
>At 08:42 AM 4/18/2001 , Karsten Riemer wrote:
>
>>Since the Specification Schema no longer has a DocumentModel, I propose to
>>remove all associated references to the analysis and structure of
>>BusinessDocuments from the Specification Schema, and leave that up to CC
>>documents as agreed in Tokyo. I think we should try to make the
>Specification
>>Schema shorter, not longer.
>