[ubl-ndrsc] RE: SQUIRTS, CCTS feedback and "Primitive Types"

["Burcham, Bill" <Bill_Burcham@stercomm.com>]

Thanks Tim.  You made my day man.  You'll have to axe Arofan re: the SQUIRT
thingie -- I think it's still a bit hush-hush.  Seriously, it was my
understanding that it was to be a separate kind of meta-object class (in
addition to e.g. BIE, ABIE, BBIE, RT) that would denote any type suitable
for use as a content component or supplementary component.  That's precisely
what "PrimitiveType" in the CCTS 1.8 feedback metamodel does/is.

I see your metamodel, and yes I see the correspondence between your {BIE,
BBIE, Data Type} and the feedback doc's {CC, BCC, PrimitiveType}.  So far,
so good.  As we stipulated in the original document -- it is immaterial
whether we speak of {CC, BCC, ACC} or {BIE, BBIE, ABIE}.

In my latest metamodel, I'm explicitly modeling content component and
supplementary components.  Turns out IMHO that UML was able to represent
that with high fidelity, what with the association roles matching up nicely
with our ISO-1197 "intuition" and all (note: I suggest we have a group
therapy session at the next F2F for anyone who has achieved an intuitive
understanding of that specification :-)

Is your "DataType" meant to encompass both the content component and
supplementary components?

I get what you're saying about the necessity of facets and rules.  I think
we could add those using a "UML Profile" (hear me Dave C?!?).  A UML Profile
lets us attach metadata to our UML Model (profiles are supported by Rose and
Argo/Poseidon UML) and let you "tunnel" data through to downstream tools.

Here's what I'm thinking now:

1. I don't care whether we model {CC,BCC...} or {BIE,BBIE...} so long as
we're consistent.  Someone please choose though!
2. I feel like the model ought to reflect a CC/BIE being comprised of a
directly associated PrimitiveType/DataType (again I don't care what we call
it -- someone please choose soon).  That directly associated object fills
the content component role relative to the CC/BIE.
3. I think the model also ought to reflect that a CC/BIE has supplementary
components and that since each of those has a (user-specified) role name
(relative to the CC/BIE) that they be associated through what we called
BCCProperty in the feedback doc (through you may want to change that name to
BBIEProperty now).
4. Also I think we should capture the XML facet information using the
standard UML Profile capability of our modeling tools.  * this (facet)
metadata may already be part of Dave Carlson's XSD UML Profile.  If not,
it's straightforward to add. * The facet metadata would be associated with
the PrimitiveType/DataType.  This brings up the issue (are you there Mike
A.?) that this UML Profile (for XSD) would become part of the UML design
deliverable for UBL.

Thoughts?

-Bill

-----Original Message-----
From: Tim McGrath [mailto:tmcgrath@portcomm.com.au]
Sent: Wednesday, July 17, 2002 8:38 PM
To: Burcham, Bill
Cc: Lisa Seaburg (E-mail); 'ubl-ndrsc@lists.oasis-open.org'
Subject: Re: SQUIRTS, CCTS feedback and "Primitive Types"


welcome back Bill!

i have also had a rethink of this model and come to a slightly different 
result (attached).  If you replace BIE with CC and Data Type with 
PrimitiveType you can compare the two.  I thought of BCC/BBIEs as an 
aggregation of PrimitiveType/DataType and possibly something else (maybe 
facets or rules for use??). what do you think?

PS what is a SQUIRT and how does it relate to Arofan (apart from the 
obvious way)?

Burcham, Bill wrote:

> There was an error in the metamodel presented in our CCTS 1.8 feedback 
> (v 5.2).  At the end of section 2 on page 7 we show BCC related to 
> Primitive Type through BCCProperty.  That was wrong because it fails 
> to distinguish the 1-1 relationship between a BCC and it's "content 
> component".  A better model is shown here (inline and attached as PDF):
>
>  
>
> <cid:046042716@17072002-0ec2>
>
>  
>
> In this model we explicitly show the distinguished relationship 
> between BCC and its content component (of meta-class "Primitive 
> Type").  The association role "contentComponent" captures this special 
> relationship explicitly.  This model retains the BCCProperty to relate 
> the BCC to its supplementary components -- each of which needs to be 
> named in relation to the BCC.
>
>  
>
> Arofan -- do I understand SQUIRTS right, that a SQUIRT is essentially 
> what I've shown as a "PrimitiveType" in the diagram.
>
>  
>
> So that's two issues:
>
> 1. we added an association directly from BCC to PrimitiveType for the 
> content component
>
> 2. we might need to change the metaclass PrimitiveType to SQUIRT
>
>  
>
> Thoughts?
>
>  
>
> -Bill
>

-- 
regards
tim mcgrath
fremantle  western australia 6160
phone: +618 93352228  fax: +618 93352142