>>>>> "Rob" == Rob Bloodgood <[EMAIL PROTECTED]> writes:
Rob> Your idea is to spawn a seperate process for running DBI that
Rob> sends an event to the parent process after a dbi call has
Rob> finished?
Yup.
Rob> Wow... that sounds so great except that in my case
Rob> it's (almost) guaranteed to fail critically as soon as I need
Rob> it to work.
Rob> <sigh>
Rob> Unless you (or anybody else) has some bulletproof, production
Rob> ready, survived-rollout! code for managing this?
This is a production ready survived rollout in use right at
this very moment solution. I don't see why you think it'd fail
critically. Forking off helper processes that you communicate with
through pipes is the tried and true Unix Way to create parallelism.
You can even isolate the code handling the DBI chunk into its own
seperate program that you can unit test independently.
--
Fletch | "If you find my answers frightening, __`'/|
[EMAIL PROTECTED] | Vincent, you should cease askin' \ o.O'
770 933-0600 x211(w) | scary questions." -- Jules =(___)=
| U
