0.1: initial skeleton
0.2: use celery for task management
0.3: basic opsdb, alarmsdb coms & test api
0.4: added some further sample resources
0.5: added meaningful alarmsdb demo method
     added snmp index to interface data to support sensu, prometheus
     increased unit test coverage to 78%
0.6: added a static demo of juniper stuff
     added some route docs to README
0.7/0.8: added static/* to release
0.9: use pyez/netconf for gathering juniper data
0.10/0.11: cache storage formatting bugfix
      logging levels configured from environment
0.12: added addresses to interface response
      put actual module number in version response
0.13: added external inventory caching
0.14: added sample route for startup-only tasks
      added method for caching selected alarmsdb tables
      added caching of last known interface status from alarmsdb
0.15: added some routes for initial classifier support
      added basic replacement for findAffectedCircuits
0.16: flatten redis storage structure
      poller api
      classifier metadata api
      read snmp community string from netconf
      derive active router list from junosspace
      read most live info from netconf
      precompute cached list of ix public & vpn rr peers
      use external logging config file
      poller api
0.17: add all interfaces to poller/interfaces response
      always use config file for celery startup
0.18: current/next database latching
0.19: added flush & latch to /jobs/update
      improved address-to-interface lookup performance
      added top-level services to responses
0.20: included both v4 & v6 addresses in peering info
0.21: added parsing of 'logical-systems' (DBOARD3-150)
0.22: return a skeleton response for unknown interfaces (DBOARD3-169)
0.23: use redis pipelines where possible
0.24: optimization, don't do aggressive pre-delete checking rebuilding
0.25: propagate errors when waiting for tasks to complete
0.26: NOT SUITABLE FOR PRODUCTION! filter qfx* routers until space is synced with opsdb