I think it would be cool to have a 'fence_email_admin.xx' fence script that
simply takes a message and email address as arguments.  It is basically a
wrapper around mail.

The idea would be to:
<method name="power">
        <device name="apc" port="2"/>
        <device name="email" msg="Node X just got fenced" email_addr="x"/>
</method>
<method name="manual">
        <device name="email" msg="Cluster foo, Node X needs manual fencing!"
email_addr="x"/>
        <device name="human" ipaddr="nodeX"/>
</method>


This fencing method would always have to return true (you wouldn't want it to
hold up the real fencing).