[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: No Subject
I would thus expect WSRM to be able to: - support a pub/sub paradigm. I believe the plumbing in the current spec is there for this: a topic is roughly equivalent to a GID. The app supplies the GID that it is publishing to, and the WSRM implementation provides the SID to ensure proper sequencing at the receiving end. But this SID is only scoped to within the domain of a single SOAP node, so messages from two different SOAP nodes cannot be ordered relative to each other. But I do not believe that they must be anyway (they aren't in JMS). But still the receiving end will require two levels of grouping to properly order the messages: (1) grouping by the "topic" that is being published to (which will be the GID), and (2) grouping by the sending node's ID so that it can order all messages arriving from that node. Thus to support the pub/sub feature, we need a "from" field also, and this must be distinct from the GID. - support for point-to-point messaging. The plumbing is essentially the same as above - the differences lie in functionality that lies outside of the WSRM-SOAP stack (like whether a message is delivered to one or more than one application). In this case a GID is equivalent to an instance of a message queue. I believe a MID may have semantic importance to an application and thus it needs control over how it is created (or at least we need to offer that option). But following the above description for how I propose that GID/SID are interpreted, the SID is generated opaquely by the WSRM provider and thus it should be kept distinct from the MID. Comments? Scott Werden WRQ > > I would appreciate more information about business situations > in which > we believe multiple un-acknowledged messages may be > outstanding. If I'm > thinking clearly, those cases in which message ordering may become > important are a subset of the un-acknowledged request cases. Those > cases in which two parties may have parallel sequences in > progress are a > further subset of this second class. We need to ensure the protocol > must go that far before we do (validate) all of the work. >
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]