Searching
  The problem with powersets
    Compared with Kademlia
  Keyword sets and network-side filtering
    Basic insertion algorithm
    Query algorithm
      Compared with other topologies (Gnutella, centralized, hybrid)
Index maintence
  Partial presence and multi-presence
    Compared against ultra-peer systems
  Meta-data gateways (supermetablockination)
    Method for aggregation of redundant network traffic
    Interaction with meta-data block expiration
      Shortish expiration time for block providers list
      Long expiration time for meta-data block
        Used for active caching
        [Should this be explained in transferring section?]
    Hysteresis of initial block insertion
      [Is this useful for anything more than making sure the right
       person suffers initial insertion?]
    Peers bump gateway, gateway bumps/updates meta-data blocks
  Lazy index replication
Transferring
  File blocking
    LBFS-like algorithm
    Unique block-sharing properties
    Two-level system (blocks compose files)
      The confusing diagram!
      [What was decided on the shallow intwined bi-directional Venti
       trees?]
  Passive caching for queue balancing
  Active caching for persistence