TODO for v0.2

• Simplify Meter and appliances: 1-to-many relationship meter:appliances. Handle dual-supply and split-mains and 3-phase mains by having multiple channels per meter.
• Maybe Meter should actually contain its appliances. i.e.:

  METER
  submeter_of: 1
  meter_type: EnviR
  room: kitchen_1
  appliances:
  - type: kettle
    instance: 1
  - type: toaster
    instance: 1
    manufacturer: blah

Advantages: don't have to specify dataset, building or meters for each appliance

• Maybe use directory structure like this:

metadata/
    dataset.yaml 
    building1/
        building1.yaml 
        appliances.yaml 
        emeter.yaml 

• which appliance is dominant on each meter
• meters can be downstream of meters in other buildings
• distinguish between "metadata shipped with dataset" and "common metadata"
• mention that meters do not have to be associated with a Building (as is the case in Tracebase)
• maybe I shouldn't bother with a JSON Schema?! Maybe I should just have a doc (based on the HDF5 Metadata doc in NILMTK) which describes the dataset; and worry about validation later?
• check that I explicitly mention where I get certain ideas (Haystack, DCI etc)
• if there's still space then add example / file hierarchy.