RE: Message adressing, intermediaries, and their description

The basic problem I see with intermediaries in a WSDL context is that
they break the basic WSDL model.

A WSDL port specifies the address of the port plus all the information
that the port receives on the wire.

At the same time, a WSDL port is used by the submitter to figure out
where to send the message (the port address) and what type of
information to put inside the message.

If there are intermediaries, then this model becomes ambiguous. Which
address are we talking about, the service's or the intermediary's. If it
is the service's, then the information in the port definition is of no
use for the submitter. If it is the intermediary's, then where is the
real port's address described?

Same problem with the wire information. Are the headers specified in the
binding the ones received by the service? Or are they the ones provided
by the submitter.

This, of course, does not mean that intermediaries always imply a
difference in address and wire information at the submitter and receiver
sides. Just that they allow it and the WSDL model currently does not
address that case.

Ugo

> -----Original Message-----
> From: www-ws-arch-request@w3.org [mailto:www-ws-arch-request@w3.org]On
> Behalf Of Hugo Haas
> Sent: Friday, October 17, 2003 6:38 AM
> To: www-ws-arch@w3.org
> Subject: Message adressing, intermediaries, and their description
> 
> 
> Following up on my message about envelope and destination address[321]
> which is I think only one part of the issue described below, which
> also relates to issue 2[322], I have tried to capture my thoughts
> underneath.
> 
> Sorry, they aren't very well structured, but I wanted to get those out
> before I disappear for a couple of weeks to get the ball rolling. The
> intermediary part is definitely related to MikeM's and MikeC's ongoing
> work. Hopefully, this email will go in the same direction.
> 
> I think that we should have intermediaries on the face-to-face
> agenda.
> 
> Mismatch between our document and SOAP 1.2:
> 
> - Our documents says that a message envelope contains address
>   information to deliver the message.
> 
> - A SOAP message, and therefore a SOAP envelope, does not contain such
>   information. It assumes that this information is known, either out
>   of band, or using an extension.
> 
> - In a SOAP message, SOAP nodes are addressed with URIs which
>   represent their roles. This role is not, in the general case, the
>   address of a SOAP node, but an abstract name: e.g.
>   "http://www.w3.org/2003/05/soap-envelope/role/next" is the next SOAP
>   node receiving the message.
> 
> Description of services and intermediaries:
> 
> [ Note that I haven't reread carefully the WSDL 1.2 specification
>   recently, so some of the things below may be wrong. Please correct
>   anything which is incorrect in light of discussions in this group. ]
> 
> - Our service description concept discusses the description of a Web
>   service's interface. The description of intermediaries is left out
>   of the picture.
> 
> - As it currently stands, WSDL 1.2 leaves out intermediaries other
>   than talking about the role attribute in the SOAP binding.
> 
> - One can wonder the role of the intermediaries in the architecture:
>   - they are visible, addressable agents (which is certainly what SOAP
>     1.2 is aiming at, I think): in this case, it may make sense to
>     have them appear in the WSDL 1.2 abstract model, not just in the
>     SOAP 1.2 binding; one way to see this is to have intermediaries
>     described as services and have a role identifier specified, and
>     then reference from other service descriptions.
>   - they are just a SOAP artifact: I know that we have talked at the
>     face-to-face meeting about talking about SOAP intermediaries as
>     being Web services intermediaries; yet, I think that this makes
>     them first-class objects.
> 
> What I think we need to do:
> 
> - In the message model, have an actor concept, which has a role
>   identified by a URI; have headers targeted to actors; say something
>   about the addressing (URL of a node) versus the targeting (URI of
>   a role that a node acts at).
> 
> - We need to figure out what kind of description of intermediaries is
>   necessary, and I think that we will need to talk to the WSDWG about
>   this, as they probably had discussions of this kind before.
> 
> I hope that this is clear and on target enough.
> 
> Regards,
> 
> Hugo
> 
>   321. 
http://lists.w3.org/Archives/Public/www-ws-arch/2003Sep/0051.html
  322. http://www.w3.org/2002/ws/arch/2/issues/wsa-issues.html#x2
-- 
Hugo Haas - W3C
mailto:hugo@w3.org - http://www.w3.org/People/Hugo/

Received on Friday, 17 October 2003 15:22:52 UTC