RE: Issue i017 - Purpose of the Action property -- my action item

Below.

> -----Original Message-----
> From: public-ws-addressing-request@w3.org [mailto:public-ws-
> addressing-request@w3.org] On Behalf Of Anish Karmarkar
> Sent: Sunday, December 19, 2004 9:59 PM
> To: public-ws-addressing@w3.org
> Subject: Issue i017 - Purpose of the Action property -- my action item
> 
> 
> During the 2004-12-13 I took an AI to send an email out to the ML regd
> issue i017.
> Going through the archives of the mailing list I see that I had
> already
> sent an email regarding this (on nov 15th). It is located at [1].
> 
> To recap that email:
> 
> 1) The [action] property is supposed to uniquely identify the
> semantics
> implied by the message. Since the value of this property is fixed by
> the
> WSDL description (either through the defaulting mechanism or through
> the
> use of wsa:Action attribute), 

More accurately, when WSDL is used to describe how a message is
constructed, that WSDL also provides facilities for specifying the
[action] property.  It is important to keep metadata separate from the
runtime.  Metadata and metadata formats may evolve.  WS-Addressing
should work well with WSDL (1.1 and 2.0), but not require services to be
built using metadata from a WSDL document in order to process messages
which use WS-Addressing.

> this value is really per message type
> within an MEP/operation/transmission primitive. Note that there are
> semantics associated with the MEP/operation grouping within an
> interface/portType as well as semantics associated with the individual
> input/output/fault message defined in WSDL. Why is it necessary to
> provide a mechanism, specifically the wsa:Action attribute, which
> overrides the default (where the default algorithm does produce a
> unique value)? 

We already debated just such a use case when we considered whether we
should have separate action defaults for WSDL 1.1 and WSDL 2.0.  Though
we decided this time that there were benefits in excess of costs in
keeping the defaults the same between 1.1 and 2.0, those tradeoffs might
change in a new version of WSDL or when a different description language
is used.

> What is the usecase for this? At the very least identical
> (wsa:Action) attribute values should be disallowed, otherwise the
> [action] property will not uniquely identify the semantics implied by
> the message (type).

The WSDL operation is (currently) a reasonable stand-in for the
"semantics of the message".  But the contract of the service might go
far beyond that, or might be captured in a form other than by making a
WSDL document available.  We should not limit the values of <wsa:Action>
to those which are conveniently mapped to WSDL.  We should instead
provide facilities to describe in WSDL the possible values that
<wsa:Action> might have.  WSDL describes the message, it doesn't
constrain what a message may be.

> 2) There is a operation name mapping requirement in WSDL 2.0 [2].
> Given
> that we have resolved issue i031 to make [action] property required, I
> see the [action] property can be used to satisfy the operation name
> mapping requirement. WS-Addressing WSDL 2.0 binding should define how
> this is done.

A proposal would be nice :-).

> HTH.
> 
> -Anish
> --
> 
> [1]
> http://lists.w3.org/Archives/Public/public-ws-
> addressing/2004Nov/0380.html
> [2] http://www.w3.org/TR/2004/WD-wsdl20-
> 20040803/#Interface_OperationName

Received on Monday, 20 December 2004 19:26:14 UTC