Re: issue-intra-port-relationship (was ..Freshly updated draft of part1 (was: Re: Overloading [was RE: Minutes, 27 June 2002 Web Service Description Telcon]))

Just to add few more cents..

I have also seen proposals limiting one service per WSDL. If that makes
through this restriction would not only restrict one from putting such
related ports not only in the same service but also in the same WSDL as
they need to be in separate services. Given the other restriction that
was proposed in the conference call that every WSDL MUST have a unique
targetNamespace, it will result in these not being in the same
targetNamespace as well..

Just want to raise the awareness regarding undesirable side-effects this
could have.

Best regards,

Prasad

Prasad Yendluri wrote:

> Hi Sanjiva (et al),
>
> Sanjiva Weerawarana wrote:
>
>> I have also closed the following issue:
>>
>>     <issue id="issue-intra-port-relationship" status="closed">
>>       <head>Should intra-port relationships be allowed?</head>
>>       <source>Prasad Yendluri</source>
>>       <p>The above restrictions seems to be unnecessary. What is the
>>
>>       justification?</p>
>>       <resolution><p>Decided to retain this restriction as no one
>> could
>>       figure out what one would want with having this feature. See
>>       Wed PM minutes for June '02 F2F.</p></resolution>
>>     </issue>
>
> Sorry I could not be at the F2F but, here is my two cents on this.
> Putting a restriction without a justification seems unreasonable to
> me. We already have several areas of confusing text in the spec that
> simply put restrictions w/o offering any explanation (e.g. only one
> part in message when 'type' AII is used; 'type' AII must be used in
> messages at abstract if SOAP binding 'use' is 'encoded' etc.).
>
> Why put restriction and tie the hands when we can't think of a reason
> for putting restriction? I offer couple of examples:
>
>    * Say I have a service that offers query and search operations that
>      are hosted on their own ports in the service. Query might need to
>      invoke the search to accomplish its purpose sometime (or vice
>      versa). Granted they can be put in separate services but if the
>      provider feels they are closely related and really belong
>      together why force one to do so?
>    * Say we have a service that offers both higher level abstraction
>      and lower-level fine-granular services (e.g. MAPI and IMAP) on
>      separate ports and the MAPI operations  may need to invoke
>      operations in IMAP port to accomplish its purpose. So the MAP
>      port would want to "communicate" with the IMAP port. Again they
>      can be in separate services and things would work fine. But these
>      are just examples and why should the spec put a restriction and
>      why not make the designers make these decisions, unless we see
>      some bad side effects in permitting this?
>
> It seems putting restrictions arbitrarily without justifying them is
> undesirable IMHO.
>
> "Ports within a service have the following relationship:
>
>    *        None of the ports communicate with each other..." ?
>
> My two cents..
>
> Regards, Prasad
>

--
---------------------------------------------------------------
Principal Architect, ATG; webMethods Inc.,
432 Lakeside Drive, Sunnyvale, CA 94088-3793, USA
Tel: (408) 962-5226 mailto: pyendluri@webmethods.com
---------------------------------------------------------------

Received on Friday, 28 June 2002 15:02:59 UTC