>I suppose that could be a memory problem, if so we could develop a
queue system using the file system and not memory, >or extend the
current queue system to use the file system when the memory queue list
is full...
>I think that a option, is to write a file to disk for each queued
message (to prevent truncate files)... the file >should also have a
predefined format (eg, receiver:message), and should be deleted when is
pushed from the queue list...

Or, a general, easily extendable and pluggable persistency manager could
be written. This way, we could have different implementations for
different needs (i.e. anything from in-memory persistance to db
support). I'd love to see a better solution to the current in-memory
queues.

- Anders


Reply via email to