RE: (Open-content model + inheritance) issues

I agree. A straightforward design would require any extension to specify
how it participates in the component model it extends. For example, an
extension to portType would specify how it applies (or not) to the
{extends} property just as it would to the other properties of portType.


By definition, extensions to the component model are part of a general
equivalence model based on the structure of the component model. It
seems less wieldy to require extensions to indicate how they contribute
(or not) to a specific equivalence function; this invites the question
about the value of a specific equivalence function like the one we have
tentatively defined for portType "inheritance".

--Jeff

-----Original Message-----
From: FABLET Youenn [mailto:fablet@crf.canon.fr] 
Sent: Friday, November 15, 2002 12:42 AM
To: www-ws-desc@w3.org
Cc: Jean-Jacques Moreau
Subject: (Open-content model + inheritance) issues

At last f2f, we agreed on an operation structural equivalence, this
equivalence being based on the am properties of the operation element.
This seems to be a simple and reasonnable solution, but all of this gets
a little trickier with the open content model, hence the following
clarification questions.

First question:
What type of inheritance do we want for open-content metadata at the
portType level ?
Second question:
How does the operation structural equivalence should take care of open
content metadata ?

First answer:
I think that the simplest solution would be to say : none.
At now this seems reasonnable because portTypes are only sets of
operations. 
If the portType contains more data than its set of operations (like
???:feature metadatas if any) then we might need something more
elaborated (for instance the extension spec says whether its metadata
are inheritable or not), but this might become quickly too complicated
IMO (?).


Second answer
At now, the consensus, from what I have understood is the following:
Rules for two operations to be equivalent :
	- their names must be the same
	- their message names must be the same


I think that we could add something like:
	- their open-content metadata that have a wsdl:required=true
must be the same (infoset-based ?)
This means that the equivalence should take into account open-content
metadata but only the "important" one (required=true).
Or should it be let to the extension authors to specify that?

Last point: what about two operations that are structurally the same but
are in two different portTypes, these portTypes having different
open-content metadata (for instance one portType has a metadata saying:
I implement only safe operations and the other has another metadata
saying: I only implement dangerous but fast & fun operations). In this
case, are the operations still the same? Should we take into account
this case in the operation structural equivalence ? At least some
clarification would be good.

Thoughts ?

	Youenn

Received on Wednesday, 20 November 2002 14:07:08 UTC