In-Multi-Out MEP [was "WSDL 2.0 specification"]

[Comments received from Shlomo Cwang <scwang@tti-telecom.com>]

-----Forwarded Message-----
From: Shlomo Cwang <scwang@tti-telecom.com>
To: 'David Booth' <dbooth@w3.org>
Subject: RE: WSDL 2.0 specification
Date: Wed, 09 Mar 2005 16:27:23 +0200

David,

By all means. Please, forward them to the public list. I hope the Working
Group will reconsider these issues.
The rationale for the In-Multi-Out MEP is the "bulk data retrieval" scenario
we very often find in Telecom and non-Telecom applications. At the back-end,
these services are usually implemented using an iterator design pattern,
which would map very nicely to the In-Multi-Out MEP. 
As for the binding, we could potentially use a CORBA IIOP or JMS binding.
Incidentally, I'm TTI's representative in the MTOSI (Multi-Technology
Operations System Interface) Working Group of the Telemanagement Forum. I'm
hoping to propose the use of WSDL 2.0 for the next phase version of the
MTOSI interface, but I need to be sure that the WSDL MEPs match the
operations' patterns of the interface.
Thanks again,

Shlomo

-----Original Message-----
From: David Booth [mailto:dbooth@w3.org]
Sent: Tuesday, March 08, 2005 9:29 PM
To: Shlomo Cwang
Subject: Re: WSDL 2.0 specification


Shlomo,

Thanks for your feedback and questions.  My answers are below.  May I
also forward them to the public list public-ws-desc-comments@w3.org for
Last Call comments, so that others can read them and the Working Group
can  track them?


On Tue, 2005-03-08 at 11:23, Shlomo Cwang wrote:
> Hi David,
> 
> My name is Shlomo and I'm investigating the potential use of WSDL 2.0 for
> the description of Telecom-based Web Services.
> I know that you're involved in the WSDL 2.0 specification and I kindly
> request your answers regarding some aspects of the spec which are not
clear
> to me. I appreciate your cooperation. The questions are:
> 
> 1) Why has the In-Multi-Out pattern been deleted from the spec?

As far as I recall there were two reasons: 1. nobody in the Working
Group had a particular need for it; 2. we don't have a binding that uses
it.

> 
> 2) We need to describe an asynchronous Request-Multiple Response
> interaction. How are we supposed to describe it without the In-Multi-Out
> pattern?

There are several options, though I don't know if you would consider any
of them good enough for your purposes.

1. Use the in-out pattern (regular request/response), but specify an
additional application-level constraint (outside of WSDL) that there may
be multiple instances of the response.  Thus, the WSDL document itself
would be insufficient for a client to make use of the service.  The
client would *also* need to know of this additional application-level
constraint.

2. Define the request and response as two separate one-way operations,
and specify application-level constraints (outside of WSDL) to indicate
that one request may be followed by multiple responses.  From a WSDL
perspective, each request would be a separate WSDL operation and each
response would be a separate operation, but the application could view
the combination of a request and multiple responses as representing a
single application-level operation.

3. Define a new MEP and corresponding binding extension for
request-multi-response.

> 
> 3) The specification states that on top of the 8 pre-defined patterns,
> additional ones may be defined, but it doesn't provide a formal
> language/notation/syntax for the pattern extensibility. How is the user of
> the spec supposed to define new ones?

Good point.  I think it would make sense to add some guidance on this to
either the spec or the primer.  The basic answer:

1. Look around on the Web to see if somebody else has already defined
one that is close enough to what you want.

2. Write an HTML document that clearly defines the MEP, and post it at a
stable URL that will represent the formal (URL) name of the MEP, such as
http://example.com/2005/ws/in-multi-out.  

3. Write a corresponding binding extension for your MEP.

4. Publicize your new MEP and binding extension, so that others can
implement and use it.

Note that the above procedure does *not* cause your MEP to become
automatically recognized and usable by WSDL toolkits.  It simply
provides a well-defined mechanism for naming and reusing them.

If you think the Working Group should reconsider its decision to drop
the request-multiple-response MEP, then please send your comments to
public-ws-desc-comments@w3.org.  Please also explain why, and describe
your use case.  


> 
> Thanks in advance,
> 
> Shlomo
> 
> 
> Shlomo Cwang 
> Communication Solutions Manager 
> TTI Telecom
> Petach Tikvah, Israel
> Office +972 3 926-9736 
> scwang@tti-telecom.com 
> 
> Information in this e-mail and its attachments is confidential and may be
> privileged. This e-mail is for the exclusive use of the intended
> recipient(s). If you are not one of the intended recipients, you are
hereby
> informed that any use, disclosure, distribution, and/or copying of this
> information is strictly prohibited. If you have received this information
in
> error, please inform the sender and then delete it from your system
> immediately. Thank you.
-- 

David Booth
W3C Fellow / Hewlett-Packard

Received on Wednesday, 9 March 2005 17:04:21 UTC