Re: [wsrm] [REL-50] {WAS Rel 44: Duplicate Elimination and Time To Live (TTL)}

[Paolo Romano <Paolo.Romano@dis.uniroma1.it>]

>    As per the un-resolved issue, sending a Fault for a pure wsdl one-way
>    operation is wrong, A Fault shouldn't be sent, unless <output> or
>     <output>/<fault> is specified (note that as per the wsdl schema you cannot
>     specify a <fault> without an <output> definition).

This is true, but you can always define a one way operation, carrying a fault.
The point is WSDL 1.1 is just not flexible enough to describe the rich MEPs a
WS-RM can enable, especially in the case of asynchronous message flows. The
choiche is among:
- not to define any WSDL description, because the next version of WSDL should
offer a solution to the current problems.
- to use BEA's approach. All messages are one-way. Faults and ACKs, in
particular, can at this point be received asynchronously. The main problem I can
see with this approach is that if an application using WS-RM was designed to
interact through a req-response MEP and it was advertized in its WSDL, then some
kind of mapping should be defined between the application WSDL and the WS-RM's
one.
- to prevent this problem, only 2 asynchronous (one-way) operations could be
defined: one to receive the ACKs and one to receive the faults. The WSDL of the
WS-RM would describe only the sintactical structure of the WS-RM headers, and
would be MEP agnostic, in the sense that no transmission primitive would be
defined at WS_RM level. It would be up to the application to define its own
MEPs, and the ws-rm processor should try to exploit the application flow of
messages in order to piggyback its acks. In the case the application was layered
over a WS-RM processor, the original application WSDL would also import the
WS-RM WSDL, or just a sub-set of it in case not all ws-rm functionalities werw
supported.

Paolo


>     May be we should use the
>     word  generate an error/exception as that doesn't always imply sending the
>     Fault to the initial sender in the SOAP response.
>
>  -Sunil




--
Paolo Romano