[Zope-DB] Large jobs and timeouts.

[Eric Kamm]

This question interests me greatly.  I'd really like to hear
what other people do.  I've struggled with this kind of
issue myself.

My current solution is to process asynchronously (fork and
detach) and provide myself way to check the status of the
separate process (maybe by having the forked process leave
a sentinel file somewhere).  The checking (polling) can be
done via HTTP refresh to enhance the web experience.

You might also want a mechanism to ensure that you don't
start the same job twice.  One solution, if the sentinel is
a file, then make the name the sentinel file as a function
of your job and its parameters.  That way, if the sentinel
file already exists, you know not to start the job again.