Re: assertion classes

Maryann Hondo wrote:
> I can't help but notice the similarity to the "usage" assertion 
> attribute in the first version of the spec[ 
> http://download.boulder.ibm.com/ibmdl/pub/software/dw/specs/ws-polfram/ws-policy2003.pdf] 
>  :-)
>
> The problem we ran into was how is this "class" or "attribute" 
>  processed in the intersection and merge operations?

I guess that would depend on the class of the assertion. It would e.g. 
make sense to completely ignore "local" assertions during merges and 
intersection while an "obligation" would be processed like any current 
non-optional assertion.

> And what processing can the framework do? and what is left to the 
> individual domains ( i.e, how do I know
> if I'm a proprietary requestor? or a proprietary provider? ( I must 
> confess, this one baffles me a bit....but maybe its like the "ignored").

Right now the only way to infer this would be from the namespace. And 
I'm thinking for "proprietary" assertions that is probably the way it 
should be. It doesn't seem to make much sense to e.g. add an additional 
attribute for that.

> Could you please elaborate on how this proposal for "classes" affects 
> the following:

This is not a proposal.

>     * attachment models ( WSDL, UDDI, etc)
>     * intersection
>     * policy authoring
>           o what is normative?
>           o what is a "guideline"?
>

I'm not sure we want to spend the time and go into that level of detail 
for now. What I wanted to show first of all is that the processing model 
and other aspects may differ between classes of assertions, and that 
e.g. optional assertions have commonalities and differences to other 
classes of assertions. I was hoping that this could streamline the 
discussion on optional a little and clarify a few common traits with 
other proposals, such as local and advisory policies/assertions.

Fabian

Received on Wednesday, 18 October 2006 01:55:10 UTC