Re: [wsrm] Duplicate of Delivered Fault revisited

[Tom Rutt <tom@coastin.com>]

Sunil Kunisetty wrote:

> Tom,
>
> Good observation. For similar reasons, I didn't want to have batching of Acks
> on R-R. Infact, for all these reasons, I never wanted (or rather was never
> enthusiastic about)  RM support for WSDL 1.1 R-R operations.
>
> So we have couple of choices here:
> 0)  Remove RM support completely for WSDL 1.1 R-R operations
>  
>
Unacceptable

> 1)  Or , say DE doesn't make sense for R-R operations
>  
>
unacceptable, this is  a main reason to use it, to protect non 
idempotent ops

> 2)  Or,  create a new thing called 'warning' (like ack and fault) and
>       for R-R DE case, deliver the msg. to the destination and send the
>       response along with the 'warning'.
>  
>
will not protect agains non idempotent operations.  Not acceptable

> 3)  Or, just send a Http response back (i.e., response doesn't have any SOAP
>      envelope or just SOAP envelope with no body/header/attachment entries.
>  
>
This will confuse the soap/wsdl processor, which is expecting a soap body.

> 4)  Or, create a new fault for DE for R-R case and send the fault...
>  
>
I believe this is the best soluction.  It is a fault condition, since 
the rmp has nothing to put in the soap body to obey the wsdl contract.


> The problem with 4 is that, if the Ack & Response was lost on the first invocation,
> he cannot ever get the response unless he changes the Message Id.
>  
>
That is ok, since we are not offering relibility on the response.

> I prefer (0), but I know it will be too drastic and critical at this stage. If not (0), I
> prefer (4).
>
> -Sunil
>
>Tom Rutt wrote:
>
>  
>
>>I have come up with a scenario, that makes me want to reconsider sending
>>an ack for a dupcate of delived message.
>>
>>Suppose we have a wsdl , non idempotent, request response operation type
>>which the user wants to protect with ws-reliability.
>>
>>Lest look at the response reply patern .
>>
>>So if the first time the operation is invoked, the receiver will deliver
>>it, and the operation response will carry the rm ack.
>>
>>Now if the sender gets nervous and resends just before it receives an
>>ack, it will be detected as duplicate, by the receiving rmp.  Now
>>the receiving rmp must not deliver this second operation invocation to
>>the receiving app, so what does it put in the soap body for
>>this response.  We are calling it a rm ack, so we will not trigger a
>>fault condition.
>>
>>What would happen if the body was empty, with no indication of faulut in
>>the ws response header.
>>
>>Perhaps we should return a "duplicateOf Delivered" fault code to convey
>>the situation in an unambiguous manner.
>>
>>--
>>----------------------------------------------------
>>Tom Rutt                email: tom@coastin.com; trutt@fsw.fujitsu.com
>>Tel: +1 732 801 5744          Fax: +1 732 774 5133
>>
>>To unsubscribe from this mailing list (and be removed from the roster of the OASIS TC), go to http://www.oasis-open.org/apps/org/workgroup/wsrm/members/leave_workgroup.php.
>>    
>>
>
>  
>

-- 
----------------------------------------------------
Tom Rutt		email: tom@coastin.com; trutt@fsw.fujitsu.com
Tel: +1 732 801 5744          Fax: +1 732 774 5133