[sc34wg3] association source

[Robert Barta]

On Wed, May 15, 2002 at 10:01:20AM +0200, Bernard Vatant wrote:
> To put on the table for XTM 2.0 maybe ...
> 
> I had this remark last week when working with an industrial customer on a TM model.
> 
> "I do not want associations in the TM that are not validated by some document"
> 
> In an industrial context, that makes sense.

Also in a 'non-industrial' context this makes sense. At least if you mean that
a somehow generated assertion...

> The TM information expressed by (or asserted
> in) associations is essentially extracted from documents (by humans or automatic tools).

...is supposed to fit into the context. 'Context' is

  - ontology (types, ...)
  - other constraints

and this is what a "TM constraint language" might deliver in some not-so-remote future.

> That led me to that kind of idea so simple that you wonder why nobody did express it
> before :))
> 
> 1. Every association is an assertion.

So far, so good.

> 2. A valid assertion should mention its source (document, author, authority ...)

Is it a

  "association #0815 was generated from document <crazy.weird> by author <the.monk>"

? This is just another association with #0815 reified:

(assertion-created-from-by)
from-document: crazy.weird
by-author: the.monk
assertion: topic-reifying-assoc-0815 

> 3. Therefore an association should (be able to) mention its source.

Voila (poor french, sorry).

> This is quite different from scope, although one should be tempted to include the source
> in a <scope> element.

I cannot be easily tempted. ;-)

> (meta)association. Too many ways in fact, when thinking about interoperability. The notion
> of source being very specific, and very important, it would be good to have a standard way
> to express it, cast in the specification stone itself.
> 
> It would be very useful and backward-compatible to include an optional <source> element
> under <association>
> The child elements of <source> being the same ones as those of <subjectIdentity>

That would be tooooo specific whereas the reification 'trick' can be applied in a
variety of ways. For instance:

    "according to a forrester research
           at the end of 2002 
              France Telecom holds 
              60% of the swedish telephone market"

The association

   (holds-market-percentage) reified by stm-1
   market: swedish-telephone-market
   share: percentage[60..]
   share-holder: france-telecom

is reified and is used simply in a qualifying statement:

   (timeframed-prediction-forecaster)
   timeframe: date[..2002Q4]
   forecaster: forrester-research
   prediction: stm-1

\rho