RE: Extensions and Import/Include

This is a good answer to my concern.  Thanks for thinking this through
so clearly!

Do you think we need to state in the spec that building a component
model is a function of the WSDL provided, plus the set of extensions the
processor understands, without mentioning the dreaded "processor" word?

> -----Original Message-----
> From: www-ws-desc-request@w3.org [mailto:www-ws-desc-request@w3.org]
On
> Behalf Of Arthur Ryman
> Sent: Thursday, June 08, 2006 5:00 AM
> To: www-ws-desc@w3.org
> Subject: Extensions and Import/Include
> 
> 
> I've had further thoughts on the interaction between extensions and
> the import/include mechanism. Jonathan asked if the WS-A extension
> would violate my proposed restrictions. I don' t think is does because
> of the recent "clarifications" about how extensions work.
> 
> I had assumed that the presence of an extension was triggered by the
> content of a document. This is wrong since we agreed that an extension
> was in effect whether or not a document actually contained any markup
> from it's namespace. For example, if you claim that a component model
> conforms to the wsdlx extension then ALL operations have a {safety}
> property whether or the wsdlx:safe attribute is present.
> 
> Therefore it is fine for WS-A to add properties to components even if
> there is no markup in those components (e.g. Interface components).
> This doesn't violate the import/include restriction because you get
> the same properties added whether or not a component is brought in via
> import/include or is defined in the document. The extension is present
> globally and uniformly throughout the component model. A processor can
> therefore read (and optionally cache) each document exactly once and
> then assemble them into the full component model instance.
> 
> This implies that we can't just talk about component models. We have
> to talk about extended component models, i.e. the core component model
> plus a set of extensions. The presence of extensions is independent of
> the document content and therefore we have to specify which extensions
> are present when we consider the validity of a component model
> instance.
> 
> -- Arthur
>

Received on Friday, 9 June 2006 18:35:24 UTC