[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [wsrm] Persistence requirement in the WS-RM specification
> <scott> > I was thinking that there would be a parameter passed as part of the WSRM > message that indicated the desired service level, e.g., > > <soap:header> > <wsrm:ReliableMessage> > <wsrm:MessageType ServiceLevel="MUST_PERSIST"> > ....... > </wsrm:MessageType> > </wsrm:ReliableMessage> > </soap:header> > > The spec would define 2 or 3 base levels of service, such as MUST_PERSIST, > DO_NOT_CARE, and MUST_DELIVER. The semantics of the latter would be (for > example): > > An ACK indicates the service level was met by the receiver. A Fault would be > available to indicate the WSRM node cannot meet the requested service level. > > Putting pre-defined service levels in the spec will allow WSRM apps to be > written once without having to write special code for each WSRM vendor that > comes along that someone wants to send messages to. > > </scott> > Seems to me to be a simple way to accomodate the need to allow devices with no persistance capabilities to use our protocol. I can't see the difference between DO_NOT_CARE and MUST_DELIVER. I think that any WS Reliable Messaging implemantion MUST always deliver the (non-duplicated and in-order) messages to the application. May be, we could cut even shorter and just prescribe two persistance levels: 1) MUST_PERSIST: whose semantic is quite obvious 2) DO_NOT_CARE : whose semantic may be the one mentioned by Scott in his mail and that I copy for reference. > "You MUST deliver the message to the application, whether that QoS > assurance is met through using persistance (to non-volitile storage) is up > to the WSRM implementation". May be, we could add a note where we claim that if no persistance is used the RM processor ONLY tolerates communication failures, but no crash failures. Paolo
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]