This is my basic netlink monitor. I've never really trusted the "persist"
option in ppp. Besides not trusting it, it's not really flexible enough for
how I'd like to tune things.

I prefer to have it try immediately on failure, then wait a bit before
trying again if it failed. I know an exponential wait time between failures
is probably better, but you run the risk of waiting too long if you have
several failed attempts in a row due to an outage. In my experience, the
probability of successfuly connecting at any time has very little to do time
that it has been down. Actualy, beyond a certain point the probablility of
success actualy increases... the "it's been down for an hour, someone's
bound to fix it soon" syndrome.

There are better ways to do this... I wrote this before I realised you could
do things like wait untill termination of a child process. If anyone actualy
does have or know of a better implementation of this, I'd appreciate being
told about it (abo@minkirri.apana.org.au). Oh, yeah, this is GPL'd so if you
modify it, I'd like to get your mods :-)

Installation

put the "netlink" script in your /etc/init.d directory (/etc/rc.d/init.d on
redhat). put "netlinkd" in /usr/local/sbin. update your runlevel stuff so
netlink is stoped/started at the right runlevels (on debian execute
"update-rc netlink defaults", on redhat "chkconfig netlink reset").

Note that the init.d script was written for debian, so it may need
modification for redhat or others. Also, the netlinkd script will probalby
need some minor tweaking for your system. In particular, the script
currently uses "pon" to restart the link, and the wait times are how _I_
like them.

Enjoy :-)