> We're already close. The linkages between SMTP, NNTP, POP3, > IMAP and the pipeline are all through the spooler or the > repositories. I've been toying with the idea of a > distributed spooler for most of the year. The problem is > that we have to synchronize access to items (only one > processor should take an item for processing), and we need to > recover if a processor fails to complete processing of an item. > > These issues are why I've started to think that instead of a > distributed spool mechanism, we should consider a central > spooler with clustered processors. It would be easier to > control and recover, since the spooler would control access, > and would know if a processor failed.
This is why I think JMS is actually a good fit for the spooler queuing mechanism. A JMS queue can have multiple readers and JMS guarantees that the each message will be delivered to one reader one time only. The reader can use manual acknowledgement of messages once it has completed processing it or secured it locally. Messages that are not acknowldeged will be delivered to another reader after a suitable timeout. Since JMS provides so much of what is needed, I think it would be a great starting point and since the spooler implementation is a configurable component it would be easy to replace later or even provide alternate implementatations for different needs. Steve --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
