RE: operation safety as semantic annotation?

Hi Paul, thanks for your comments, here's my replies. 8-)

On Mon, 2007-01-29 at 15:34 +0000, paul.downey@bt.com wrote:
> > Operation safety is, at least to me, a clear semantic annotation. It
> > says nothing about the structure of the interface, 
> 
> but it can be used to determine if a binding can use HTTP GET, which isn't just semantic ..

It's a defaulting behavior based on what the operation promises not to
do, because HTTP GET is not just URI there and something back, it's also
that obligation on what it won't do. GET has a semantic requirement that
the operations need to fulfill. 

In any case, that part of the spec that does the HTTP GET defaulting
would not be affected, as it uses the {safety} property, not the
wsdlx:safe attribute.

> 
> > I would propose that we change the syntax from wsdlx:safe="true"  to
> > sawsdl:modelReference="http://www.w3.org/2006/01/wsdl-extensions#SafeInteraction"
> 
> what's the issue with the current short syntax - can't sawsdl easily define a mapping?

The issue is that my semantic client deals with various aspects of the
operations regarding what they do (incl. safety), and it will need to
know both about sawsdl:modelReferences and the particular WSDL extension
called wsdlx:safe.

Of course a mapping can be defined, but since WSDL and SAWSDL are so
close in their respective processes, we should consider doing this one
thing using one mechanism. In fact I guess I'm asking the WSDL WG here
to consider doing the mapping on the WSDL side. 8-)

> do you think that's likely to encourage more people to flag their 
> "getCustomerAddress" method as being safe, something I think we're 
> both after?

Good catch, can't guarantee that my approach would get more adoption of
the flag, and I could see how early people experimenting with it could
be irritated by the more complex syntax and one more spec involved.

I believe the adoption of safety annotations really hinges on the
availability (or lack) of applications that would make use of it. 
I'm making one such application and I'm saying that the current syntax
is a bit of an impediment. 8-)

Hope it makes sense,
Jacek

Received on Monday, 29 January 2007 16:10:13 UTC