TODO

==Preparing for distributed-ness

The main idea behind distributed Sadie is a bunch of Sadie servers that share an nfs volume and which make use of a Redis cluster.  So I need to be careful that no single sadie server need be "special".

As it stands, the session has these things to consider:

* mutexes
* expiry data structure
* refresh data structure
* registered keys

For mutexes, I'm thinking a locking abstraction will do.  Something like:

  lock_id = acquire_lock( params )
  release_lock( lock_id )

Once abstracted, this should be relatively straightforward to extend as needed for distributed-ness.

[ NOTE: lock abstraction is now done ]  

The expiry and refresh data structures are currently implemented as red-black trees, but it would be ideal if there were a way to share this information amongst the sadie instances.  I'm thinking it might make sense to only a single sadie instance to be expiring at a time (and then only for a set amount of time).  Further, I'm thinking refresh could be synchronized such that only a single sadie instance can be finding a refreshable key at a time, but, once it's selected one, it releases its lock so that another instance can begin a search for a another refreshable.

[ NOTE: the expiry and refresh data structures are now maintained by a 'timestamp_queue' class that will make it easier to swap in redis-based functionality ]

The registered keys data structure can actually be maintained separately in each instance in as much as the primers should really be the same across the instances.