RE: Issue 140 bogus?

 Stuart,
 I take back calling your issue bogus and non-issue. You and
Chris have shown me that there are issues behind #140 that need
to be addressed.
 I didn't mean to insult you, Stuart, I only read the text of
#140 and that was what I understood.
 Now on to how we can solve the issues. 8-)

 I'd like to split issue #140 to
Issue 140a: boolean amIThatActor(URI actorURI, Context context) :
	what should be in the context?
Issue 140b: routing the message (related at least to issue #103)

 As for issue 140a: my experience in implementing SOAP leads me
to saying that this should be implementation-dependent. We could
just say:
 "A node assumes the roles of a set of actors, identified by
their respective Actor URIs, constant for this message."
 (This wouldn't prevent an implementation or an extension from
setting this set dynamically even during the message processing,
it would only require that when the message leaves the node, the
set can be known and the message looks as though processed by
that set of actors. My double-quoted sentence might need some
tweaking to reflect this note.)
 So my take on the function's interface is:
 boolean amIThatActor(URI actorURI, Everything theWorld)

 As for issue 140b, I think we decided that a source-routing
extension is out of scope of the XMLP WG. I think the spec should
say that the "where-to-send-the-message" must be known *somehow*.
The "somehow" part depends on message patterns, contracts, WSDL
(et al) and routing extensions. 8-)

 Hoping to have split the issue appropriately,

                            Jacek Kopecky

                            Idoox
                            http://www.idoox.com/



On Tue, 2 Oct 2001, Williams, Stuart wrote:

 > Hi Jacek,
 >
 > I think that the matter is a little different. I guess that we could take
 > two different tacks, here.
 >
 > The tack you seem to be following is that the default actor is just another
 > actor, and the issue of determining whether you are a given actor is no
 > different for the default actor than it is for an other actor. If we persue
 > that tack, then I think I would continue to have an issue in that our
 > current draft then offers little if any discussion of possible basis for a
 > SOAP node to determine that it performs the role of any given actor
 > (particularly if the roles its willing to play wrt to a given message are
 > dependent on the content of the message - which is partially hinted at).
 >
 > The other tack which I was persuing was the the anonymous actor, the "actor
 > who cannot be named" (think Harry Potter :-), is infact different. All SOAP
 > messages have one. On the surface at least there is nothing in the message
 > (other than the message itself) that gives a basis for a Node to determine
 > whether it acts in the default role for a given message. For non-default
 > actors the actor notion is at least evocative of the simple idea the the
 > message should visit (in some unspecified order) a set of actors directly
 > referenced by the value of the associated actor attribute. For this simple
 > case, resolution of the actor name to a request-URI would at least apprear
 > obvious for http based actor URIs and using the http binding (no-doubt there
 > are some subtlties). Resolving
 > http://www.w3.org/2001/09/soap-envelope/actor/default this way isn't very
 > helpful!
 >
 > Issue 140 is that there is little/no discussion of these matters in the
 > draft spec, narrowly cast as a problem in resolving whether a SOAP Node is
 > the default actor with respect to a given message. I think your response
 > would cause me to want to broaden the issue rather than regard it as bogus!
 >
 > On the constructive side, I've been thinking about this a bit as well. I
 > think this takes us down a path of questioning the relationship between the
 > request-URI (an HTTP binding concept)/transport endpoints(maybe more
 > generic), a SOAP node, the actor roles supported by a SOAP Node (possibly
 > wrt to a given message), and our notions of what sort of things we might be
 > naming/addressing as SOAP endpoints.
 >
 > I think a pragmatic way out might be to acknowledge that the choice of
 > transport endpoint for a given message (or message exchange) may be the
 > thing that determines the set of actor roles engaged at that physical node
 > to process an inbound message. One could concieve of the following
 > functions:
 >
 > 1)	boolean IAmThatActor(ActorUri);
 >
 > 2)	boolean IAmThatActor(ActorUri,Message);
 >
 > 3)	boolean IAmThatActor(ActorUri,InboundTransportEndpoint);
 >
 > 4)	boolean IAmThatActor(ActorUri,Message,InboundTransportEndpoint);
 >
 >
 > 1) is akin to the simple scheme mentioned above, and is not very useful when
 > it comes to the default/anon actor. 2) is a little more useful, and would
 > admit behaviour like SOAP-RP or ebXML where header fields (targetted at
 > ../next) will influence the determination. 3) basically captures a notion of
 > the a collection of actors being configured behind a given transport
 > endpoint (and useful intermediaries would likely not list default/anon
 > amongst the actors at the endpoints they use :-)). 4) is a hybrid, that
 > seems to have all the bells and whistles (apart from time dependent
 > variation, the phase of the moon or whatever).
 >
 > Sorry, this is so long... its the first time I've written this down. I think
 > basically I am optimistic that we could resolve Issue 140. I don't think its
 > a non-Issue. I don't think that the spec currently gives much guidance about
 > the self determination by a SOAP Node of whether it acts in a particular
 > role with respect to a given message. That said, I don't think that requires
 > lots of discussion.
 >
 > Regards
 >
 > Stuart
 >
 >
 >
 > > -----Original Message-----
 > > From: Jacek Kopecky [mailto:jacek@idoox.com]
 > > Sent: 02 October 2001 17:45
 > > To: Stuart Williams
 > > Cc: David Fallside; xml-dist-app@w3.org
 > > Subject: Issue 140 bogus?
 > >
 > >
 > >  Stuart,
 > >  I believe your issue with the section 2.3 of the first part of
 > > our draft spec is not an issue.
 > >  The actor URI, if present, is not specified to be anything in
 > > particular, it's just a URI that the targetted actor somehow
 > > knows is point at him. If instead of targetting the default actor
 > > with a missing actor URI we targetted it with the URI
 > > "http://www.w3.org/2001/09/soap-envelope/actor/default", the
 > > situation would be completely equivalent.
 > >  And yes, the set of Actor URIs a node acts as is an *outside*
 > > information to the envelope.
 > >  Best regards,
 > >
 > >                             Jacek Kopecky
 > >
 > >                             Idoox
 > >                             http://www.idoox.com/
 > >
 >

Received on Tuesday, 2 October 2001 14:59:46 UTC