issue 26 discussion summary

Hi all,

this email attempts to summarize our discussions on issue 26 [1].

There are three questions in this issue:
     1. Does interface categorization (or any modelReference) apply on
        operation?
     2. Add categorization on operation level?
     3. Need precedence rules for categorization on interface and
        operation?

We seem to like the idea of categorization on the operation level, so we
probably can add in section 2.1.2 some text similar to 2.1.1 talking
about categorization, and pointing to appendix D. With this addition, we
would be talking about using modelReference on WSDL operation to
categorize it or to "provide a high level description of the operation".

However, due to the potential complexity, we don't seem to want to make
categorization on interface to apply on the operations as well, as this
would either make all interface modelReferences propagate to the
operations, or we would have to introduce a hard notion of
categorization, which we eschewed so far.

However, as we can envision that interface category may in fact apply to
the operations as well, we can note in our specifications that the
creator of the categorization model (e.g. something based on UNSPSC) can
specify that if a category is attached to a WSDL interface using
sawsdl:modelReference, this category will propagate to the operations as
well.

A processor that does not know this particular category will not see it
on the operations, but that should be OK, since the processor cannot
really do anything much with this modelReference anyway. A processor
that does know the category will behave like it was on all the
operations as well, because it will obey the propagation rule specified
by the category model.

Finally, on the question of precedence rules, I believe that our "all
apply" handling of modelReferences is sufficient, and of course, the
specifications for the particular semantic models (e.g. categories) can
say whatever is appropriate about the relations (e.g. precedence)
withing the model or to other semantic models.

To summarize, it seems we could close issue 26 by
      * adding operation categorization text in 2.1.2
      * mentioning somewhere that semantic models (e.g. categorization
        taxonomies) may specify propagation of their annotations from
        interface to operations, in which case they may want to handle
        precedences as well

Hope it helps,

Jacek

[1] http://www.w3.org/2002/ws/sawsdl/issues/#x26

Received on Saturday, 19 August 2006 11:38:32 UTC