You should have a separate webapp that can push a new configuration 
file to Horde. Anything else that let an end user go into Horde itself 
and hit a button would either need a separate config - again, shared 
across a cluster - and/or would probably be brought down by the same 
issue you were trying to use the button for (permissions checking, ... 
etc.).

Also, you can set 'horde' to 'inactive' in the registry.

Good point re: the database.  But what other mechanism could one use 
across a cluster?



With regards to a postauthenticate hook, yes that'd work.  Basically I 
was looking for something that could be modified from the GUI (like 
alarms) so that it didn't have to be an admin who was intimately 
familiar with the horde code who would have to do it.  If I'm not 
around when things go kablooie we need to be able to have a 
straightforward way close the doors.  At the moment I go into our 
firewall and redirect connections to a static "we're broke" message 
but that's not a particularly elegant solution.

Aside from reading it from the database, what would this do that you 
couldn't do now with a postauthenticate hook that just returned false? 
And if you do put it in the database, how will it help you when the 
database is down?

It would be enormously handy to have a way to prevent logins across 
the board.  For example if the mailservers, memcache servers, or 
database go down we'd like to be able to flip a switch and all logins 
would be temporarily disabled until we fix the problem.  It would be 
also helpful to be able to set a notice message at the same time which 
would notify users of the system problems.  This would need to be 
stored in the database to be able to affect all machines in the cluster.