Version 1.4.0 Release

 * [maybe] make the "top" link point to a list of service groups, linked to
   pages that just display that one section (as opposed to the view that
   "running" gives, which is all sections)
 * examine the benefits of schedulers having write control over host data
   . all that information is connection-oriented, and thogh the schedulers need
     that info, it is not part of their purpose: determining where the current
     request should be directed
   . connection information should be at the proxy level (the TCP server
     that opens connections to the proxied host)
   . proxies should track information for each of its connections (really, the
     connections being made to the proxy's receiver)
 * examine breaking out listeners and managers across different processes/hosts
   . if a Twisted service is to be loadbalanced how should that happen?
   . if we are running a manger daemon for twisted services that are configured
     to be loadbalanced, how should that happen?
   . what about all of the services potentially being equal? any can be a
     manager, any can be balanced via another services proxy manager
 * sticky connections. keep a cache of 'client_ip':'backend_ip'
   . manager needs to be able to call "scheduler.cleanup()" to allow the
     cache to be manageable.
 * continue adding unit tests