[OT] Dilemma about OS <-> Postgres interaction

I have an application, largely written in plpgsql, where messages are
dumped into an 'outbox' for sending via email or text message.

I am in a dilemma about how best to pick up these messages and send them
on to the message sending systems. At present I have looping perl
scripts which have a permanent database connection, which I run through
screen sessions on my server. Rather embarassingly whenever I upgrade my
application code (reload the database and export the latest CVS code) or
upgrade Postgres, my looping scripts die.

I suppose I could use plperl and create a trigger to alert a system-wide
daemon that a message was ready for sending whenever a new one dropped
into the 'outbox', and restart the daemon if it had died. I feel
superstitious about using an untrusted language for functions though.

Any sage advice? I realise this is both a beginner's quandry, and also
not particularly Postgresql specific. Apologies.

Rory
--
Rory Campbell-Lange
<rory(at)campbell-lange(dot)net>
<www.campbell-lange.net>