<br>Ben and Paul, thanks for your thoughts.<br><br>My producer is generating a constant stream of messages with keys Type.Id, where the Type is one of a few hundred possible values and the Id identifies is an entity where something of that Type just happened. More concretely, in a trouble ticketing system for example, tickets have an Id and they constantly get updated with stuff of a certain Type.<br>
<br>After a particular Type of update, I need to start processing all following updates on the ticket with the given Id, up until some kind of "stop" update happens.<br><br>So I was thinking of having one process per ticket that needs monitored (~1000 at any one time). But the producer doesn't know which ones will get monitored. And I have to ignore all messages that are not of type "start monitoring" and aren't currently monitored by a process. Thus I can't use the mandatory flag either.<br>
<br>I'll try playing with the "look and forward" approach to see how awkward it'll be.<br><br>Thanks,<br>Robby<br><br>