Multi-hop Ad Hoc Wireless Communication
draft-baccelli-intarea-adhoc-wireless-com-01

Endpoint          a 32-bit opaque and locally unique value, which
  identifier        identifies a particular endpoint of a particular
                    DNCP node. The value 0 is reserved for DNCP and
                    DNCP-based protocol purposes and not used to
                    identify an actual endpoint. This definition is in
                    sync with the interface index definition in
                    [RFC3493], as the non-zero small positive integers
                    should comfortably fit within 32 bits.

  Peer              another DNCP node with which a DNCP node
                    communicates using at least one particular local
                    and remote endpoint pair.

  Node data        a set of TLVs published and owned by a node in the
                    DNCP network. Other nodes pass it along as-is, even
                    if they cannot fully interpret it.

  Origination Time  the (estimated) time when the node data set with
                    the current sequence number was published.

  Node state        a set of metadata attributes for node data. It
                    includes a sequence number for versioning, a hash
                    value for comparing equality of stored node data,
                    and a timestamp indicating the time passed since
                    its last publication (i.e., since the origination
                    time). The hash function and the length of the hash
                    value are defined in the DNCP profile.

  Network state    a hash value which represents the current state of
  hash              the network.  The hash function and the length of
                    the hash value are defined in the DNCP profile.
                    Whenever a node is added, removed or updates its
                    published node data this hash value changes as
                    well.  For calculation, please see Section 4.1.

  Trust verdict    a statement about the trustworthiness of a
                    certificate announced by a node participating in
                    the certificate based trust consensus mechanism.

  Effective trust  the trust verdict with the highest priority within
  verdict          the set of trust verdicts announced for the
                    certificate in the DNCP network.

  Topology graph    the undirected graph of DNCP nodes produced by
                    retaining only bidirectional peer relationships
                    between nodes.

  Bidirectionally  a peer is locally unidirectionally reachable if a

Stenberg & Barth          Expires May 5, 2016                  [Page 7]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

  reachable        consistent multicast or any unicast DNCP message
                    has been received by the local node (see Section
                    4.5).  If said peer in return also considers the
                    local node unidirectionally reachable, then
                    bidirectionally reachability is established.  As
                    this process is based on publishing peer
                    relationships and evaluating the resulting topology
                    graph as described in Section 4.6, this information
                    is available to the whole DNCP network.

  Trickle Instance  a distinct Trickle [RFC6206] algorithm state kept
                    by a node (Section 5) and related to an endpoint or
                    a particular (peer, endpoint) tuple with Trickle
                    variables I, t and c. See Section 4.3.

2.1.  Requirements Language

  The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
  "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
  "OPTIONAL" in this document are to be interpreted as described in RFC
  2119 [RFC2119].

3.  Overview

  DNCP operates primarily using unicast exchanges between nodes, and
  may use multicast for Trickle-based shared state dissemination and
  topology discovery.  If used in pure unicast mode with unreliable
  transport, Trickle is also used between peers.

  DNCP is based on exchanging TLVs (Section 7) and defines a set of
  mandatory and optional ones for its operation.  They are categorized
  into TLVs for requesting information (Section 7.1), transmitting data
  (Section 7.2) and being published as data (Section 7.3).  DNCP based
  protocols usually specify additional ones to extend the capabilities.

  DNCP discovers the topology of the nodes in the DNCP network and
  maintains the liveliness of published node data by ensuring that the
  publishing node is bidirectionally reachable.  New potential peers
  can be discovered autonomously on multicast-enabled links, their
  addresses may be manually configured or they may be found by some
  other means defined in the particular DNCP profile.  The DNCP profile
  may specify, for example, a well-known anycast address or
  provisioning the remote address to contact via some other protocol
  such as DHCPv6 [RFC3315].

  A hash tree of height 1, rooted in itself, is maintained by each node
  to represent the state of all currently reachable nodes (see
  Section 4.1) and the Trickle algorithm is used to trigger

Stenberg & Barth          Expires May 5, 2016                  [Page 8]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

  synchronization (see Section 4.3).  The need to check peer nodes for
  state changes is thereby determined by comparing the current root of
  their respective hash trees, i.e., their individually calculated
  network state hashes.

  Before joining a DNCP network, a node starts with a hash tree that
  has only one leaf if the node publishes some TLVs, and no leaves
  otherwise.  It then announces the network state hash calculated from
  the hash tree by means of the Trickle algorithm on all its configured
  endpoints.

  When an update is detected by a node (e.g., by receiving a different
  network state hash from a peer) the originator of the event is
  requested to provide a list of the state of all nodes, i.e., all the
  information it uses to calculate its own hash tree.  The node uses
  the list to determine whether its own information is outdated and -
  if necessary - requests the actual node data that has changed.

  Whenever a node's local copy of any node data and its hash tree are
  updated (e.g., due to its own or another node's node state changing
  or due to a peer being added or removed) its Trickle instances are
  reset which eventually causes any update to be propagated to all of
  its peers.

4.  Operation

4.1.  Hash Tree

  Each DNCP node maintains an arbitrary width hash tree of height 1.
  The root of the tree represents the overall network state hash and is
  used to determine whether the view of the network of two or more
  nodes is consistent and shared.  Each leaf represents one
  bidirectionally reachable DNCP node.  Every time a node is added or
  removed from the topology graph (Section 4.6) it is likewise added or
  removed as a leaf.  At any time the leaves of the tree are ordered in
  ascending order of the node identifiers of the nodes they represent.

4.1.1.  Calculating network state and node data hashes

  The network state hash and the node data hashes are calculated using
  the hash function defined in the DNCP profile (Section 9) and
  truncated to the number of bits specified therein.

  Individual node data hashes are calculated by applying the function
  and truncation on the respective node's node data as published in the
  Node State TLV.  Such node data sets are always ordered as defined in
  Section 7.2.3.

Stenberg & Barth          Expires May 5, 2016                  [Page 9]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

  The network state hash is calculated by applying the function and
  truncation on the concatenated network state.  This state is formed
  by first concatenating each node's sequence number (in network byte
  order) with its node data hash to form a per-node datum for each
  node.  These per-node data are then concatenated in ascending order
  of the respective node's node identifier, i.e., in the order that the
  nodes appear in the hash tree.

4.1.2.  Updating network state and node data hashes

  The network state hash and the node data hashes are updated on-demand
  and whenever any locally stored per-node state changes.  This
  includes local unidirectional reachability encoded in the published
  Peer TLVs (Section 7.3.1) and - when combined with remote data -
  results in awareness of bidirectional reachability changes.

4.2.  Data Transport

  DNCP has few requirements for the underlying transport; it requires
  some way of transmitting either unicast datagram or stream data to a
  peer and, if used in multicast mode, a way of sending multicast
  datagrams.  As multicast is used only to identify potential new DNCP
  nodes and to send status messages which merely notify that a unicast
  exchange should be triggered, the multicast transport does not have
  to be secured.  If unicast security is desired and one of the built-
  in security methods is to be used, support for some TLS-derived
  transport scheme - such as TLS [RFC5246] on top of TCP or DTLS
  [RFC6347] on top of UDP - is also required.  They provide for
  integrity protection and confidentiality of the node data, as well as
  authentication and authorization using the schemes defined in
  Security and Trust Management (Section 8).  A specific definition of
  the transport(s) in use and their parameters MUST be provided by the
  DNCP profile.

  TLVs (Section 7) are sent across the transport as is, and they SHOULD
  be sent together where, e.g., MTU considerations do not recommend
  sending them in multiple batches.  DNCP does not fragment or
  reassemble TLVs thus it MUST be ensured that the underlying transport
  performs these operations should they be necessary.  If this document
  indicates sending one or more TLVs, then the sending node does not
  need to keep track of the packets sent after handing them over to the
  respective transport, i.e., reliable DNCP operation is ensured merely
  by the explicitly defined timers and state machines such as Trickle
  (Section 4.3).  TLVs in general are handled individually and
  statelessly (and thus do not need to be sent in any particular order)
  with one exception: To form bidirectional peer relationships DNCP
  requires identification of the endpoints used for communication.  As
  bidirectional peer relationships are required for validating

Stenberg & Barth          Expires May 5, 2016                [Page 10]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

  liveliness of published node data as described in Section 4.6, a DNCP
  node MUST send a Node Endpoint TLV (Section 7.2.1).  When it is sent
  varies, depending on the underlying transport, but conceptually it
  should be available whenever processing a Network State TLV:

  o  If using a stream transport, the TLV MUST be sent at least once
      per connection, but SHOULD NOT be sent more than once.

  o  If using a datagram transport, it MUST be included in every
      datagram that also contains a Network State TLV (Section 7.2.2)
      and MUST be located before any such TLV.  It SHOULD also be
      included in any other datagram, to speed up initial peer
      detection.

  Given the assorted transport options as well as potential endpoint
  configuration, a DNCP endpoint may be used in various transport
  modes:

  Unicast:

      *  If only reliable unicast transport is used, Trickle is not used
        at all.  Whenever the locally calculated network state hash
        changes, a single Network State TLV (Section 7.2.2) is sent to
        every unicast peer.  Additionally, recently changed Node State
        TLVs (Section 7.2.3) MAY be included.

      *  If only unreliable unicast transport is used, Trickle state is
        kept per peer and it is used to send Network State TLVs
        intermittently, as specified in Section 4.3.

  Multicast+Unicast:  If multicast datagram transport is available on
      an endpoint, Trickle state is only maintained for the endpoint as
      a whole.  It is used to send Network State TLVs periodically, as
      specified in Section 4.3.  Additionally, per-endpoint keep-alives
      MAY be defined in the DNCP profile, as specified in Section 6.1.2.

  MulticastListen+Unicast:  Just like Unicast, except multicast
      transmissions are listened to in order to detect changes of the
      highest node identifier.  This mode is used only if the DNCP
      profile supports dense multicast-enabled link optimization
      (Section 6.2).

4.3.  Trickle-Driven Status Updates

  The Trickle algorithm [RFC6206] is used to ensure protocol
  reliability over unreliable multicast or unicast transports.  For
  reliable unicast transports, its actual algorithm is unnecessary and
  omitted (Section 4.2).  DNCP maintains multiple Trickle states as

Stenberg & Barth          Expires May 5, 2016                [Page 11]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

  defined in Section 5.  Each such state can be based on different
  parameters (see below) and is responsible for ensuring that a
  specific peer or all peers on the respective endpoint are regularly
  provided with the node's current locally calculated network state
  hash for state comparison, i.e., to detect potential divergence in
  the perceived network state.

  Trickle defines 3 parameters: Imin, Imax and k.  Imin and Imax
  represent the minimum value for I and the maximum number of doublings
  of Imin, where I is the time interval during which at least k Trickle
  updates must be seen on an endpoint to prevent local state
  transmission.  The actual suggested Trickle algorithm parameters are
  DNCP profile specific, as described in Section 9.

  The Trickle state for all Trickle instances defined in Section 5 is
  considered inconsistent and reset if and only if the locally
  calculated network state hash changes.  This occurs either due to a
  change in the local node's own node data, or due to receipt of more
  recent data from another node as explained in Section 4.1.  A node
  MUST NOT reset its Trickle state merely based on receiving a Network
  State TLV (Section 7.2.2) with a network state hash which is
  different from its locally calculated one.

  Every time a particular Trickle instance indicates that an update
  should be sent, the node MUST send a Network State TLV
  (Section 7.2.2) if and only if:

  o  the endpoint is in Multicast+Unicast transport mode, in which case
      the TLV MUST be sent over multicast.

  o  the endpoint is NOT in Multicast+Unicast transport mode, and the
      unicast transport is unreliable, in which case the TLV MUST be
      sent over unicast.

  A (sub)set of all Node State TLVs (Section 7.2.3) MAY also be
  included, unless it is defined as undesirable for some reason by the
  DNCP profile, or to avoid exposure of the node state TLVs by
  transmitting them within insecure multicast when using secure
  unicast.

4.4.  Processing of Received TLVs

  This section describes how received TLVs are processed.  The DNCP
  profile may specify when to ignore particular TLVs, e.g., to modify
  security properties - see Section 9 for what may be safely defined to
  be ignored in a profile.  Any 'reply' mentioned in the steps below
  denotes sending of the specified TLV(s) to the originator of the TLV
  being processed.  All such replies MUST be sent using unicast.  If

Stenberg & Barth          Expires May 5, 2016                [Page 12]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

  the TLV being replied to was received via multicast and it was sent
  to a multiple access link, the reply MUST be delayed by a random
  timespan in [0, Imin/2], to avoid potential simultaneous replies that
  may cause problems on some links, unless specified differently in the
  DNCP profile.  Sending of replies MAY also be rate-limited or omitted
  for a short period of time by an implementation.  However, if the TLV
  is not forbidden by the DNCP profile, an implementation MUST reply to
  retransmissions of the TLV with a non-zero probability to avoid
  starvation which would break the state synchronization.

  A DNCP node MUST process TLVs received from any valid (e.g.,
  correctly scoped) address, as specified by the DNCP profile and the
  configuration of a particular endpoint, whether this address is known
  to be the address of a peer or not.  This provision satisfies the
  needs of monitoring or other host software that needs to discover the
  DNCP topology without adding to the state in the network.

  Upon receipt of:

  o  Request Network State TLV (Section 7.1.1): The receiver MUST reply
      with a Network State TLV (Section 7.2.2) and a Node State TLV
      (Section 7.2.3) for each node data used to calculate the network
      state hash.  The Node State TLVs SHOULD NOT contain the optional
      node data part to avoid redundant transmission of node data,
      unless explicitly specified in the DNCP profile.

  o  Request Node State TLV (Section 7.1.2): If the receiver has node
      data for the corresponding node, it MUST reply with a Node State
      TLV (Section 7.2.3) for the corresponding node.  The optional node
      data part MUST be included in the TLV.

  o  Network State TLV (Section 7.2.2): If the network state hash
      differs from the locally calculated network state hash, and the
      receiver is unaware of any particular node state differences with
      the sender, the receiver MUST reply with a Request Network State
      TLV (Section 7.1.1).  These replies MUST be rate limited to only
      at most one reply per link per unique network state hash within
      Imin.  The simplest way to ensure this rate limit is a timestamp
      indicating requests, and sending at most one Request Network State
      TLV (Section 7.1.1) per Imin.  To facilitate faster state
      synchronization, if a Request Network State TLV is sent in a
      reply, a local, current Network State TLV MAY also be sent.

  o  Node State TLV (Section 7.2.3):

      *  If the node identifier matches the local node identifier and
        the TLV has a greater sequence number than its current local
        value, or the same sequence number and a different hash, the

Stenberg & Barth          Expires May 5, 2016                [Page 13]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

        node SHOULD re-publish its own node data with a sequence number
        significantly (e.g., 1000) greater than the received one, to
        reclaim the node identifier.  This difference is needed in
        order to ensure that it is higher than any potentially
        lingering copies of the node state in the network.  This may
        occur normally once due to the local node restarting and not
        storing the most recently used sequence number.  If this occurs
        more than once or for nodes not re-publishing their own node
        data, the DNCP profile MUST provide guidance on how to handle
        these situations as it indicates the existence of another
        active node with the same node identifier.

      *  If the node identifier does not match the local node
        identifier, and one or more of the following conditions are
        true:

        +  The local information is outdated for the corresponding node
            (local sequence number is less than that within the TLV).

        +  The local information is potentially incorrect (local
            sequence number matches but the node data hash differs).

        +  There is no data for that node altogether.

        Then:

        +  If the TLV contains the Node Data field, it SHOULD also be
            verified by ensuring that the locally calculated hash of the
            Node Data matches the content of the H(Node Data) field
            within the TLV.  If they differ, the TLV SHOULD be ignored
            and not processed further.

        +  If the TLV does not contain the Node Data field, and the
            H(Node Data) field within the TLV differs from the local
            node data hash for that node (or there is none), the
            receiver MUST reply with a Request Node State TLV
            (Section 7.1.2) for the corresponding node.

        +  Otherwise the receiver MUST update its locally stored state
            for that node (node data based on Node Data field if
            present, sequence number and relative time) to match the
            received TLV.

      For comparison purposes of the sequence number, a looping
      comparison function MUST be used to avoid problems in case of
      overflow.  The comparison function a < b <=> ((a - b) % (2^32)) &
      (2^31) != 0 where (a % b) represents the remainder of a modulo b

Stenberg & Barth          Expires May 5, 2016                [Page 14]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

      and (a & b) represents bitwise conjunction of a and b is
      RECOMMENDED unless the DNCP profile defines another.

  o  Any other TLV: TLVs not recognized by the receiver MUST be
      silently ignored unless they are sent within another TLV (for
      example, TLVs within the Node Data field of a Node State TLV).
      TLVs within the Node Data field of the Node State TLV not
      recognized by the receiver MUST be retained for distribution to
      other nodes and for calculating the node data hash as described in
      Section 7.2.3 but are ignored for other purposes.

  If secure unicast transport is configured for an endpoint, any Node
  State TLVs received over insecure multicast MUST be silently ignored.

4.5.  Discovering, Adding and Removing Peers

  Peer relations are established between neighbors using one or more
  mutually connected endpoints.  Such neighbors exchange information
  about network state and published data directly and through
  transitivity this information then propagates throughout the network.

  New peers are discovered using the regular unicast or multicast
  transport defined in the DNCP profile (Section 9).  This process is
  not distinguished from peer addition, i.e., an unknown peer is simply
  discovered by receiving regular DNCP protocol TLVs from it and
  dedicated discovery messages or TLVs do not exist.  For unicast-only
  transports, the individual node's transport addresses are
  preconfigured or obtained using an external service discovery
  protocol.  In the presence of a multicast transport, messages from
  unknown peers are handled in the same way as multicast messages from
  peers that are already known, thus new peers are simply discovered
  when sending their regular DNCP protocol TLVs using multicast.

  When receiving a Node Endpoint TLV (Section 7.2.1) on an endpoint
  from an unknown peer:

  o  If received over unicast, the remote node MUST be added as a peer
      on the endpoint and a Peer TLV (Section 7.3.1) MUST be created for
      it.

  o  If received over multicast, the node MAY be sent a (possibly rate-
      limited) unicast Request Network State TLV (Section 7.1.1).

  If keep-alives specified in Section 6.1 are NOT sent by the peer
  (either the DNCP profile does not specify the use of keep-alives or
  the particular peer chooses not to send keep-alives), some other
  existing local transport-specific means (such as Ethernet carrier-
  detection or TCP keep-alive) MUST be used to ensure its presence.  If

Stenberg & Barth          Expires May 5, 2016                [Page 15]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

  the peer does not send keep-alives, and no means to verify presence
  of the peer are available, the peer MUST be considered no longer
  present and it SHOULD NOT be added back as a peer until it starts
  sending keep-alives again.  When the peer is no longer present, the
  Peer TLV and the local DNCP peer state MUST be removed.  DNCP does
  not define an explicit message or TLV for indicating the termination
  of DNCP operation by the terminating node, however a derived protocol
  could specify an extension, if the need arises.

  If the local endpoint is in the Multicast-Listen+Unicast transport
  mode, a Peer TLV (Section 7.3.1) MUST NOT be published for the peers
  not having the highest node identifier.

4.6.  Data Liveliness Validation

  Maintenance of the hash tree (Section 4.1) and thereby network state
  hash updates depend on up-to-date information on bidirectional node
  reachability derived from the contents of a topology graph.  This
  graph changes whenever nodes are added to or removed from the network
  or when bidirectional connectivity between existing nodes is
  established or lost.  Therefore the graph MUST be updated either
  immediately or with a small delay shorter than the DNCP profile-
  defined Trickle Imin, whenever:

  o  A Peer TLV or a whole node is added or removed, or

  o  the origination time (in milliseconds) of some node's node data is
      less than current time - 2^32 + 2^15.

  The artificial upper limit for the origination time is used to
  gracefully avoid overflows of the origination time and allow for the
  node to republish its data as noted in Section 7.2.3.

  The topology graph update starts with the local node marked as
  reachable and all other nodes marked as unreachable.  Other nodes are
  then iteratively marked as reachable using the following algorithm: A
  candidate not-yet-reachable node N with an endpoint NE is marked as
  reachable if there is a reachable node R with an endpoint RE that
  meet all of the following criteria:

  o  The origination time (in milliseconds) of R's node data is greater
      than current time - 2^32 + 2^15.

  o  R publishes a Peer TLV with:

      *  Peer Node Identifier = N's node identifier

      *  Peer Endpoint Identifier = NE's endpoint identifier

Stenberg & Barth          Expires May 5, 2016                [Page 16]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

      *  Endpoint Identifier = RE's endpoint identifier

  o  N publishes a Peer TLV with:

      *  Peer Node Identifier = R's node identifier

      *  Peer Endpoint Identifier = RE's endpoint identifier

      *  Endpoint Identifier = NE's endpoint identifier

  The algorithm terminates, when no more candidate nodes fulfilling
  these criteria can be found.

  DNCP nodes that have not been reachable in the most recent topology
  graph traversal MUST NOT be used for calculation of the network state
  hash, be provided to any applications that need to use the whole TLV
  graph, or be provided to remote nodes.  They MAY be forgotten
  immediately after the topology graph traversal, however it is
  RECOMMENDED to keep them at least briefly to improve the speed of
  DNCP network state convergence.  This reduces the number of queries
  needed to reconverge during both initial network convergence and when
  a part of the network loses and regains bidirectional connectivity
  within that time period.

5.  Data Model

  This section describes the local data structures a minimal
  implementation might use.  This section is provided only as a
  convenience for the implementor.  Some of the optional extensions
  (Section 6) describe additional data requirements, and some optional
  parts of the core protocol may also require more.

  A DNCP node has:

  o  A data structure containing data about the most recently sent
      Request Network State TLVs (Section 7.1.1).  The simplest option
      is keeping a timestamp of the most recent request (required to
      fulfill reply rate limiting specified in Section 4.4).

  A DNCP node has for every DNCP node in the DNCP network:

  o  Node identifier: the unique identifier of the node.  The length,
      how it is produced, and how collisions are handled, is up to the
      DNCP profile.

  o  Node data: the set of TLV tuples published by that particular
      node.  As they are transmitted ordered (see Node State TLV

Stenberg & Barth          Expires May 5, 2016                [Page 17]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

      (Section 7.2.3) for details), maintaining the order within the
      data structure here may be reasonable.

  o  Latest sequence number: the 32-bit sequence number that is
      incremented any time the TLV set is published.  The comparison
      function used to compare them is described in Section 4.4.

  o  Origination time: the (estimated) time when the current TLV set
      with the current sequence number was published.  It is used to
      populate the Milliseconds Since Origination field in a Node State
      TLV (Section 7.2.3).  Ideally it also has millisecond accuracy.

  Additionally, a DNCP node has a set of endpoints for which DNCP is
  configured to be used.  For each such endpoint, a node has:

  o  Endpoint identifier: the 32-bit opaque locally unique value
      identifying the endpoint within a node.  It SHOULD NOT be reused
      immediately after an endpoint is disabled.

  o  Trickle instance: the endpoint's Trickle instance with parameters
      I, T, and c (only on an endpoint in Multicast+Unicast transport
      mode).

  and one (or more) of the following:

  o  Interface: the assigned local network interface.

  o  Unicast address: the DNCP node it should connect with.

  o  Set of addresses: the DNCP nodes from which connections are
      accepted.

  For each remote (peer, endpoint) pair detected on a local endpoint, a
  DNCP node has:

  o  Node identifier: the unique identifier of the peer.

  o  Endpoint identifier: the unique endpoint identifier used by the
      peer.

  o  Peer address: the most recently used address of the peer
      (authenticated and authorized, if security is enabled).

  o  Trickle instance: the particular peer's Trickle instance with
      parameters I, T, and c (only on an endpoint in Unicast mode, when
      using an unreliable unicast transport) .

Stenberg & Barth          Expires May 5, 2016                [Page 18]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

6.  Optional Extensions

  This section specifies extensions to the core protocol that a DNCP
  profile may specify to be used.

6.1.  Keep-Alives

  While DNCP provides mechanisms for discovery and adding of new peers
  on an endpoint (Section 4.5), as well as state change notifications,
  another mechanism may be needed to get rid of old, no longer valid
  peers if the transport or lower layers do not provide one as noted in
  Section 4.6.

  If keep-alives are not specified in the DNCP profile, the rest of
  this subsection MUST be ignored.

  A DNCP profile MAY specify either per-endpoint (sent using multicast
  to all DNCP nodes connected to a multicast-enabled link) or per-peer
  (sent using unicast to each peer individually) keep-alive support.

  For every endpoint that a keep-alive is specified for in the DNCP
  profile, the endpoint-specific keep-alive interval MUST be
  maintained.  By default, it is DNCP_KEEPALIVE_INTERVAL.  If there is
  a local value that is preferred for that for any reason
  (configuration, energy conservation, media type, ..), it can be
  substituted instead.  If a non-default keep-alive interval is used on
  any endpoint, a DNCP node MUST publish appropriate Keep-Alive
  Interval TLV(s) (Section 7.3.2) within its node data.

6.1.1.  Data Model Additions

  The following additions to the Data Model (Section 5) are needed to
  support keep-alives:

  For each configured endpoint that has per-endpoint keep-alives
  enabled:

  o  Last sent: If a timestamp which indicates the last time a Network
      State TLV (Section 7.2.2) was sent over that interface.

  For each remote (peer, endpoint) pair detected on a local endpoint, a
  DNCP node has:

  o  Last contact timestamp: a timestamp which indicates the last time
      a consistent Network State TLV (Section 7.2.2) was received from
      the peer over multicast, or anything was received over unicast.
      Failing to update it for a certain amount of time as specified in

Stenberg & Barth          Expires May 5, 2016                [Page 19]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

      Section 6.1.5 results in the removal of the peer.  When adding a
      new peer, it is initialized to the current time.

  o  Last sent: If per-peer keep-alives are enabled, a timestamp which
      indicates the last time a Network State TLV (Section 7.2.2) was
      sent to to that point-to-point peer.  When adding a new peer, it
      is initialized to the current time.

6.1.2.  Per-Endpoint Periodic Keep-Alives

  If per-endpoint keep-alives are enabled on an endpoint in
  Multicast+Unicast transport mode, and if no traffic containing a
  Network State TLV (Section 7.2.2) has been sent to a particular
  endpoint within the endpoint-specific keep-alive interval, a Network
  State TLV (Section 7.2.2) MUST be sent on that endpoint, and a new
  Trickle interval started, as specified in the step 2 of Section 4.2
  of [RFC6206].  The actual sending time SHOULD be further delayed by a
  random timespan in [0, Imin/2].

6.1.3.  Per-Peer Periodic Keep-Alives

  If per-peer keep-alives are enabled on a unicast-only endpoint, and
  if no traffic containing a Network State TLV (Section 7.2.2) has been
  sent to a particular peer within the endpoint-specific keep-alive
  interval, a Network State TLV (Section 7.2.2) MUST be sent to the
  peer, and a new Trickle interval started, as specified in the step 2
  of Section 4.2 of [RFC6206].

6.1.4.  Received TLV Processing Additions

  If a TLV is received over unicast from the peer, the Last contact
  timestamp for the peer MUST be updated.

  On receipt of a Network State TLV (Section 7.2.2) which is consistent
  with the locally calculated network state hash, the Last contact
  timestamp for the peer MUST be updated in order to maintain it as a
  peer.

6.1.5.  Peer Removal

  For every peer on every endpoint, the endpoint-specific keep-alive
  interval must be calculated by looking for Keep-Alive Interval TLVs
  (Section 7.3.2) published by the node, and if none exist, using the
  default value of DNCP_KEEPALIVE_INTERVAL.  If the peer's Last contact
  timestamp has not been updated for at least locally chosen
  potentially endpoint-specific keep-alive multiplier (defaults to
  DNCP_KEEPALIVE_MULTIPLIER) times the peer's endpoint-specific keep-

Stenberg & Barth          Expires May 5, 2016                [Page 20]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

  alive interval, the Peer TLV for that peer and the local DNCP peer
  state MUST be removed.

6.2.  Support For Dense Multicast-Enabled Links

  This optimization is needed to avoid a state space explosion.  Given
  a large set of DNCP nodes publishing data on an endpoint that uses
  multicast on a link, every node will add a Peer TLV (Section 7.3.1)
  for each peer.  While Trickle limits the amount of traffic on the
  link in stable state to some extent, the total amount of data that is
  added to and maintained in the DNCP network given N nodes on a
  multicast-enabled link is O(N^2).  Additionally if per-peer keep-
  alives are used, there will be O(N^2) keep-alives running on the link
  if liveliness of peers is not ensured using some other way (e.g., TCP
  connection lifetime, layer 2 notification, per-endpoint keep-alive).

  An upper bound for the number of peers that are allowed for a
  particular type of link that an endpoint in Multicast+Unicast
  transport mode is used on SHOULD be provided by a DNCP profile, but
  MAY also be chosen at runtime.  The main consideration when selecting
  a bound (if any) for a particular type of link should be whether it
  supports multicast traffic, and whether a too large number of peers
  case is likely to happen during the use of that DNCP profile on that
  particular type of link.  If neither is likely, there is little point
  specifying support for this for that particular link type.

  If a DNCP profile does not support this extension at all, the rest of
  this subsection MUST be ignored.  This is because when this extension
  is used, the state within the DNCP network only contains a subset of
  the full topology of the network.  Therefore every node must be aware
  of the potential of it being used in a particular DNCP profile.

  If the specified upper bound is exceeded for some endpoint in
  Multicast+Unicast transport mode and if the node does not have the
  highest node identifier on the link, it SHOULD treat the endpoint as
  a unicast endpoint connected to the node that has the highest node
  identifier detected on the link, therefore transitioning to
  Multicast-listen+Unicast transport mode.  See Section 4.2 for
  implications on the specific endpoint behavior.  The nodes in
  Multicast-listen+Unicast transport mode MUST keep listening to
  multicast traffic to both receive messages from the node(s) still in
  Multicast+Unicast mode, and as well to react to nodes with a greater
  node identifier appearing.  If the highest node identifier present on
  the link changes, the remote unicast address of the endpoints in
  Multicast-Listen+Unicast transport mode MUST be changed.  If the node
  identifier of the local node is the highest one, the node MUST switch
  back to, or stay in Multicast+Unicast mode, and form peer
  relationships with all peers as specified in Section 4.5.

Stenberg & Barth          Expires May 5, 2016                [Page 21]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

7.  Type-Length-Value Objects

  0                  1                  2                  3
  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |            Type              |          Length              |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |              Value (if any) (+padding (if any))              |
  ..
  |                    (variable # of bytes)                    |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                    (Optional nested TLVs)                    |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  Each TLV is encoded as:

  o  a 2 byte Type field

  o  a 2 byte Length field which contains the length of the Value field
      in bytes; 0 means no Value

  o  the Value itself (if any)

  o  padding bytes with value of zero up to the next 4 byte boundary if
      the Length is not divisible by 4.

  While padding bytes MUST NOT be included in the number stored in the
  Length field of the TLV, if the TLV is enclosed within another TLV,
  then the padding is included in the enclosing TLV's Length value.

  Each TLV which does not define optional fields or variable-length
  content MAY be sent with additional sub-TLVs appended after the TLV
  to allow for extensibility.  When handling such TLV types, each node
  MUST accept received TLVs that are longer than the fixed fields
  specified for the particular type, and ignore the sub-TLVs with
  either unknown types, or not supported within that particular TLV
  type.  If any sub-TLVs are present, the Length field of the TLV
  describes the number of bytes from the first byte of the TLV's own
  Value (if any) to the last (padding) byte of the last sub-TLV.

  For example, type=123 (0x7b) TLV with value 'x' (120 = 0x78) is
  encoded as: 007B 0001 7800 0000.  If it were to have sub-TLV of
  type=124 (0x7c) with value 'y', it would be encoded as 007B 000C 7800
  0000 007C 0001 7900 0000.

  In this section, the following special notation is used:

      .. = octet string concatenation operation.

Stenberg & Barth          Expires May 5, 2016                [Page 22]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

      H(x) = non-cryptographic hash function specified by DNCP profile.

7.1.  Request TLVs

7.1.1.  Request Network State TLV

  0                  1                  2                  3
  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |  Type: REQ-NETWORK-STATE (1)  |          Length: >= 0        |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  This TLV is used to request response with a Network State TLV
  (Section 7.2.2) and all Node State TLVs (Section 7.2.3) (without node
  data).

7.1.2.  Request Node State TLV

  0                  1                  2                  3
  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |    Type: REQ-NODE-STATE (2)  |          Length: > 0          |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                        Node Identifier                        |
  |                  (length fixed in DNCP profile)              |
  ...
  |                                                              |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  This TLV is used to request a Node State TLV (Section 7.2.3)
  (including node data) for the node with the matching node identifier.

7.2.  Data TLVs

7.2.1.  Node Endpoint TLV

  0                  1                  2                  3
  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |  Type: NODE-ENDPOINT (3)    |          Length: > 4          |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                        Node Identifier                        |
  |                  (length fixed in DNCP profile)              |
  ...
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                      Endpoint Identifier                      |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

Stenberg & Barth          Expires May 5, 2016                [Page 23]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

  This TLV identifies both the local node's node identifier, as well as
  the particular endpoint's endpoint identifier.  Section 4.2 specifies
  when it is sent.

7.2.2.  Network State TLV

  0                  1                  2                  3
  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |    Type: NETWORK-STATE (4)    |          Length: > 0          |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |    H(sequence number of node 1 .. H(node data of node 1) ..  |
  |    .. sequence number of node N .. H(node data of node N))    |
  |                  (length fixed in DNCP profile)              |
  ...
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  This TLV contains the current network state hash calculated by its
  sender (Section 4.1 describes the algorithm).

7.2.3.  Node State TLV

  0                  1                  2                  3
  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |      Type: NODE-STATE (5)    |          Length: > 8          |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                        Node Identifier                        |
  |                  (length fixed in DNCP profile)              |
  ...
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                      Sequence Number                        |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                Milliseconds Since Origination                |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                        H(Node Data)                          |
  |                  (length fixed in DNCP profile)              |
  ...
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |      (optionally) Node Data (a set of nested TLVs)          |
  ...
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  This TLV represents the local node's knowledge about the published
  state of a node in the DNCP network identified by the Node Identifier
  field in the TLV.

Stenberg & Barth          Expires May 5, 2016                [Page 24]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

  Every node, including the node publishing the node data, MUST update
  the Milliseconds Since Origination whenever it sends a Node State TLV
  based on when the node estimates the data was originally published.
  This is, e.g., to ensure that any relative timestamps contained
  within the published node data can be correctly offset and
  interpreted.  Ultimately, what is provided is just an approximation,
  as transmission delays are not accounted for.

  Absent any changes, if the originating node notices that the 32-bit
  milliseconds since origination value would be close to overflow
  (greater than 2^32-2^16), the node MUST re-publish its TLVs even if
  there is no change.  In other words, absent any other changes, the
  TLV set MUST be re-published roughly every 48 days.

  The actual node data of the node may be included within the TLV as
  well in the optional Node Data field.  The set of TLVs MUST be
  strictly ordered based on ascending binary content (including TLV
  type and length).  This enables, e.g., efficient state delta
  processing and no-copy indexing by TLV type by the recipient.  The
  Node Data content MUST be passed along exactly as it was received.
  It SHOULD be also verified on receipt that the locally calculated
  H(Node Data) matches the content of the field within the TLV, and if
  the hash differs, the TLV SHOULD be ignored.

7.3.  Data TLVs within Node State TLV

  These TLVs are published by the DNCP nodes, and therefore only
  encoded in the Node Data field of Node State TLVs.  If encountered
  outside Node State TLV, they MUST be silently ignored.

7.3.1.  Peer TLV

  0                  1                  2                  3
  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |      Type: PEER (8)          |          Length: > 8          |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                      Peer Node Identifier                    |
  |                  (length fixed in DNCP profile)              |
  ...
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                    Peer Endpoint Identifier                  |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                  (Local) Endpoint Identifier                |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  This TLV indicates that the node in question vouches that the
  specified peer is reachable by it on the specified local endpoint.

Stenberg & Barth          Expires May 5, 2016                [Page 25]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

  The presence of this TLV at least guarantees that the node publishing
  it has received traffic from the peer recently.  For guaranteed up-
  to-date bidirectional reachability, the existence of both nodes'
  matching Peer TLVs needs to be checked.

7.3.2.  Keep-Alive Interval TLV

  0                  1                  2                  3
  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  | Type: KEEP-ALIVE-INTERVAL (9) |          Length: >= 8        |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                      Endpoint Identifier                      |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                          Interval                            |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  This TLV indicates a non-default interval being used to send keep-
  alives specified in Section 6.1.

  Endpoint identifier is used to identify the particular (local)
  endpoint for which the interval applies on the sending node.  If 0,
  it applies for ALL endpoints for which no specific TLV exists.

  Interval specifies the interval in milliseconds at which the node
  sends keep-alives.  A value of zero means no keep-alives are sent at
  all; in that case, some lower layer mechanism that ensures presence
  of nodes MUST be available and used.

8.  Security and Trust Management

  If specified in the DNCP profile, either DTLS [RFC6347] or TLS
  [RFC5246] may be used to authenticate and encrypt either some (if
  specified optional in the profile), or all unicast traffic.  The
  following methods for establishing trust are defined, but it is up to
  the DNCP profile to specify which ones may, should or must be
  supported.

8.1.  Pre-Shared Key Based Trust Method

  A PSK-based trust model is a simple security management mechanism
  that allows an administrator to deploy devices to an existing network
  by configuring them with a pre-defined key, similar to the
  configuration of an administrator password or WPA-key.  Although
  limited in nature it is useful to provide a user-friendly security
  mechanism for smaller networks.

Stenberg & Barth          Expires May 5, 2016                [Page 26]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

8.2.  PKI Based Trust Method

  A PKI-based trust-model enables more advanced management capabilities
  at the cost of increased complexity and bootstrapping effort.  It
  however allows trust to be managed in a centralized manner and is
  therefore useful for larger networks with a need for an authoritative
  trust management.

8.3.  Certificate Based Trust Consensus Method

  For some scenarios - such as bootstrapping a mostly unmanaged network
  - the methods described above may not provide a desirable tradeoff
  between security and user experience.  This section includes guidance
  for implementing an opportunistic security [RFC7435] method which
  DNCP profiles can build upon and adapt for their specific
  requirements.

  The certificate-based consensus model is designed to be a compromise
  between trust management effort and flexibility.  It is based on
  X.509-certificates and allows each DNCP node to provide a trust
  verdict on any other certificate and a consensus is found to
  determine whether a node using this certificate or any certificate
  signed by it is to be trusted.

  A DNCP node not using this security method MUST ignore all announced
  trust verdicts and MUST NOT announce any such verdicts by itself,
  i.e., any other normative language in this subsection does not apply
  to it.

  The current effective trust verdict for any certificate is defined as
  the one with the highest priority from all trust verdicts announced
  for said certificate at the time.

8.3.1.  Trust Verdicts

  Trust verdicts are statements of DNCP nodes about the trustworthiness
  of X.509-certificates.  There are 5 possible trust verdicts in order
  of ascending priority:

      0 (Neutral): no trust verdict exists but the DNCP network should
      determine one.

      1 (Cached Trust): the last known effective trust verdict was
      Configured or Cached Trust.

      2 (Cached Distrust): the last known effective trust verdict was
      Configured or Cached Distrust.

Stenberg & Barth          Expires May 5, 2016                [Page 27]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

      3 (Configured Trust): trustworthy based upon an external ceremony
      or configuration.

      4 (Configured Distrust): not trustworthy based upon an external
      ceremony or configuration.

  Trust verdicts are differentiated in 3 groups:

  o  Configured verdicts are used to announce explicit trust verdicts a
      node has based on any external trust bootstrap or predefined
      relation a node has formed with a given certificate.

  o  Cached verdicts are used to retain the last known trust state in
      case all nodes with configured verdicts about a given certificate
      have been disconnected or turned off.

  o  The Neutral verdict is used to announce a new node intending to
      join the network so a final verdict for it can be found.

  The current effective trust verdict for any certificate is defined as
  the one with the highest priority within the set of trust verdicts
  announced for the certificate in the DNCP network.  A node MUST be
  trusted for participating in the DNCP network if and only if the
  current effective trust verdict for its own certificate or any one in
  its certificate hierarchy is (Cached or Configured) Trust and none of
  the certificates in its hierarchy have an effective trust verdict of
  (Cached or Configured) Distrust.  In case a node has a configured
  verdict, which is different from the current effective trust verdict
  for a certificate, the current effective trust verdict takes
  precedence in deciding trustworthiness.  Despite that, the node still
  retains and announces its configured verdict.

8.3.2.  Trust Cache

  Each node SHOULD maintain a trust cache containing the current
  effective trust verdicts for all certificates currently announced in
  the DNCP network.  This cache is used as a backup of the last known
  state in case there is no node announcing a configured verdict for a
  known certificate.  It SHOULD be saved to a non-volatile memory at
  reasonable time intervals to survive a reboot or power outage.

  Every time a node (re)joins the network or detects the change of an
  effective trust verdict for any certificate, it will synchronize its
  cache, i.e., store new effective trust verdicts overwriting any
  previously cached verdicts.  Configured verdicts are stored in the
  cache as their respective cached counterparts.  Neutral verdicts are
  never stored and do not override existing cached verdicts.

Stenberg & Barth          Expires May 5, 2016                [Page 28]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

8.3.3.  Announcement of Verdicts

  A node SHOULD always announce any configured trust verdicts it has
  established by itself, and it MUST do so if announcing the configured
  trust verdict leads to a change in the current effective trust
  verdict for the respective certificate.  In absence of configured
  verdicts, it MUST announce cached trust verdicts it has stored in its
  trust cache, if one of the following conditions applies:

  o  The stored trust verdict is Cached Trust and the current effective
      trust verdict for the certificate is Neutral or does not exist.

  o  The stored trust verdict is Cached Distrust and the current
      effective trust verdict for the certificate is Cached Trust.

  A node rechecks these conditions whenever it detects changes of
  announced trust verdicts anywhere in the network.

  Upon encountering a node with a hierarchy of certificates for which
  there is no effective trust verdict, a node adds a Neutral Trust-
  Verdict-TLV to its node data for all certificates found in the
  hierarchy, and publishes it until an effective trust verdict
  different from Neutral can be found for any of the certificates, or a
  reasonable amount of time (10 minutes is suggested) with no reaction
  and no further authentication attempts has passed.  Such trust
  verdicts SHOULD also be limited in rate and number to prevent denial-
  of-service attacks.

  Trust verdicts are announced using Trust-Verdict TLVs:

  0                  1                  2                  3
  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |  Type: Trust-Verdict (10)    |        Length: > 36          |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |    Verdict    |                (reserved)                    |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                                                              |
  |                                                              |
  |                                                              |
  |                      SHA-256 Fingerprint                      |
  |                                                              |
  |                                                              |
  |                                                              |
  |                                                              |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                          Common Name                          |

Stenberg & Barth          Expires May 5, 2016                [Page 29]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

      Verdict represents the numerical index of the trust verdict.

      (reserved) is reserved for future additions and MUST be set to 0
      when creating TLVs and ignored when parsing them.

      SHA-256 Fingerprint contains the SHA-256 [RFC6234] hash value of
      the certificate in DER-format.

      Common Name contains the variable-length (1-64 bytes) common name
      of the certificate.

8.3.4.  Bootstrap Ceremonies

  The following non-exhaustive list of methods describes possible ways
  to establish trust relationships between DNCP nodes and node
  certificates.  Trust establishment is a two-way process in which the
  existing network must trust the newly added node and the newly added
  node must trust at least one of its peer nodes.  It is therefore
  necessary that both the newly added node and an already trusted node
  perform such a ceremony to successfully introduce a node into the
  DNCP network.  In all cases an administrator MUST be provided with
  external means to identify the node belonging to a certificate based
  on its fingerprint and a meaningful common name.

8.3.4.1.  Trust by Identification

  A node implementing certificate-based trust MUST provide an interface
  to retrieve the current set of effective trust verdicts, fingerprints
  and names of all certificates currently known and set configured
  trust verdicts to be announced.  Alternatively it MAY provide a
  companion DNCP node or application with these capabilities with which
  it has a pre-established trust relationship.

8.3.4.2.  Preconfigured Trust

  A node MAY be preconfigured to trust a certain set of node or CA
  certificates.  However such trust relationships MUST NOT result in
  unwanted or unrelated trust for nodes not intended to be run inside
  the same network (e.g., all other devices by the same manufacturer).

8.3.4.3.  Trust on Button Press

  A node MAY provide a physical or virtual interface to put one or more
  of its internal network interfaces temporarily into a mode in which
  it trusts the certificate of the first DNCP node it can successfully
  establish a connection with.

Stenberg & Barth          Expires May 5, 2016                [Page 30]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

8.3.4.4.  Trust on First Use

  A node which is not associated with any other DNCP node MAY trust the
  certificate of the first DNCP node it can successfully establish a
  connection with.  This method MUST NOT be used when the node has
  already associated with any other DNCP node.

9.  DNCP Profile-Specific Definitions

  Each DNCP profile MUST specify the following aspects:

  o  Unicast and optionally multicast transport protocol(s) to be used.
      If multicast-based node and status discovery is desired, a
      datagram-based transport supporting multicast has to be available.

  o  How the chosen transport(s) are secured: Not at all, optionally or
      always with the TLS scheme defined here using one or more of the
      methods, or with something else.  If the links with DNCP nodes can
      be sufficiently secured or isolated, it is possible to run DNCP in
      a secure manner without using any form of authentication or
      encryption.

  o  Transport protocols' parameters such as port numbers to be used,
      or multicast address to be used.  Unicast, multicast, and secure
      unicast may each require different parameters, if applicable.

  o  When receiving TLVs, what sort of TLVs are ignored in addition -
      as specified in Section 4.4 - e.g., for security reasons.  While
      the security of the node data published within the Node State TLVs
      is already ensured by the base specification (if secure mode is
      enabled, Node State TLVs are sent only via unicast as multicast
      ones are ignored on receipt), if a profile adds TLVs that are sent
      outside the node data, a profile should indicate whether or not
      those TLVs should be ignored if they are received via multicast or
      non-secured unicast.  A DNCP profile may define the following DNCP
      TLVs to be safely ignored:

      *  Anything received over multicast, except Node Endpoint TLV
        (Section 7.2.1) and Network State TLV (Section 7.2.2).

      *  Any TLVs received over unreliable unicast or multicast at too
        high rate; Trickle will ensure eventual convergence given the
        rate slows down at some point.

  o  How to deal with node identifier collision as described in
      Section 4.4.  Main options are either for one or both nodes to
      assign new node identifiers to themselves, or to notify someone
      about a fatal error condition in the DNCP network.

Stenberg & Barth          Expires May 5, 2016                [Page 31]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

  o  Imin, Imax and k ranges to be suggested for implementations to be
      used in the Trickle algorithm.  The Trickle algorithm does not
      require these to be the same across all implementations for it to
      work, but similar orders of magnitude helps implementations of a
      DNCP profile to behave more consistently and to facilitate
      estimation of lower and upper bounds for convergence behavior of
      the network.

  o  Hash function H(x) to be used, and how many bits of the output are
      actually used.  The chosen hash function is used to handle both
      hashing of node data, and to produce network state hash, which is
      a hash of node data hashes.  SHA-256 defined in [RFC6234] is the
      recommended default choice, but a non-cryptographic hash function
      could be used as well.  If there is a hash collision in the
      network state hash, the network will effectively be partitioned to
      partitions that believe that they are up to date, but actually no
      longer converged.  The network will converge either when some node
      data anywhere in the network changes, or when conflicting Node
      State TLVs get transmitted across the partition (either caused by
      Trickle-Driven Status Updates (Section 4.3) or as part of the
      Processing of Received TLVs (Section 4.4)).  If a node publishes
      node data with a hash that collides with any previously published
      node data, the update may not be (fully) propagated and the old
      version of node data may be used instead.

  o  DNCP_NODE_IDENTIFIER_LENGTH: The fixed length of a node identifier
      (in bytes).

  o  Whether to send keep-alives, and if so, whether per-endpoint
      (requires multicast transport), or per-peer.  Keep-alive has also
      associated parameters:

      *  DNCP_KEEPALIVE_INTERVAL: How often keep-alives are to be sent
        by default (if enabled).

      *  DNCP_KEEPALIVE_MULTIPLIER: How many times the
        DNCP_KEEPALIVE_INTERVAL (or peer-supplied keep-alive interval
        value) a node may not be heard from to be considered still
        valid.  This is just a default used in absence of any other
        configuration information, or particular per-endpoint
        configuration.

  o  Whether to support dense multicast-enabled link optimization
      (Section 6.2) or not.

  For some guidance on choosing transport and security options, please
  see Appendix B.

Stenberg & Barth          Expires May 5, 2016                [Page 32]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

10.  Security Considerations

  DNCP-based protocols may use multicast to indicate DNCP state changes
  and for keep-alive purposes.  However, no actual published data TLVs
  will be sent across that channel.  Therefore an attacker may only
  learn hash values of the state within DNCP and may be able to trigger
  unicast synchronization attempts between nodes on a local link this
  way.  A DNCP node MUST therefore rate-limit its reactions to
  multicast packets.

  When using DNCP to bootstrap a network, PKI based solutions may have
  issues when validating certificates due to potentially unavailable
  accurate time, or due to inability to use the network to either check
  Certificate Revocation Lists or perform on-line validation.

  The Certificate-based trust consensus mechanism defined in this
  document allows for a consenting revocation, however in case of a
  compromised device the trust cache may be poisoned before the actual
  revocation happens allowing the distrusted device to rejoin the
  network using a different identity.  Stopping such an attack might
  require physical intervention and flushing of the trust caches.

11.  IANA Considerations

  IANA should set up a registry for the (decimal 16-bit) "DNCP TLV
  Types" under "Distributed Node Consensus Protocol (DNCP)", with the
  following initial contents: ([RFC Editor: please remove] ideally as
  http://www.iana.org/assignments/dncp-registry)

      0: Reserved

      1: Request network state

      2: Request node state

      3: Node endpoint

      4: Network state

      5: Node state

      6: Reserved (was: Custom)

      7: Reserved (was: Fragment count)

      8: Peer

      9: Keep-alive interval

Stenberg & Barth          Expires May 5, 2016                [Page 33]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

      10: Trust-Verdict

      11-31: Free - policy of standards action [RFC5226] should be used

      32-511: Reserved for per-DNCP profile use

      512-767: Free - policy of standards action [RFC5226] should be
      used

      768-1023: Private use [RFC5226]

      1024-65535: Reserved for future protocol evolution (for example,
      DNCP version 2)

12.  References

12.1.  Normative references

  [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/
              RFC2119, March 1997,
              <http://www.rfc-editor.org/info/rfc2119>.

  [RFC6206]  Levis, P., Clausen, T., Hui, J., Gnawali, O., and J. Ko,
              "The Trickle Algorithm", RFC 6206, DOI 10.17487/RFC6206,
              March 2011, <http://www.rfc-editor.org/info/rfc6206>.

  [RFC6234]  Eastlake 3rd, D. and T. Hansen, "US Secure Hash Algorithms
              (SHA and SHA-based HMAC and HKDF)", RFC 6234, DOI
              10.17487/RFC6234, May 2011,
              <http://www.rfc-editor.org/info/rfc6234>.

  [RFC5226]  Narten, T. and H. Alvestrand, "Guidelines for Writing an
              IANA Considerations Section in RFCs", BCP 26, RFC 5226,
              DOI 10.17487/RFC5226, May 2008,
              <http://www.rfc-editor.org/info/rfc5226>.

12.2.  Informative references

  [RFC3493]  Gilligan, R., Thomson, S., Bound, J., McCann, J., and W.
              Stevens, "Basic Socket Interface Extensions for IPv6", RFC
              3493, DOI 10.17487/RFC3493, February 2003,
              <http://www.rfc-editor.org/info/rfc3493>.

  [RFC3315]  Droms, R., Ed., Bound, J., Volz, B., Lemon, T., Perkins,
              C., and M. Carney, "Dynamic Host Configuration Protocol
              for IPv6 (DHCPv6)", RFC 3315, DOI 10.17487/RFC3315, July
              2003, <http://www.rfc-editor.org/info/rfc3315>.

Stenberg & Barth          Expires May 5, 2016                [Page 34]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

  [RFC6347]  Rescorla, E. and N. Modadugu, "Datagram Transport Layer
              Security Version 1.2", RFC 6347, DOI 10.17487/RFC6347,
              January 2012, <http://www.rfc-editor.org/info/rfc6347>.

  [RFC5246]  Dierks, T. and E. Rescorla, "The Transport Layer Security
              (TLS) Protocol Version 1.2", RFC 5246, DOI 10.17487/
              RFC5246, August 2008,
              <http://www.rfc-editor.org/info/rfc5246>.

  [RFC7435]  Dukhovni, V., "Opportunistic Security: Some Protection
              Most of the Time", RFC 7435, DOI 10.17487/RFC7435,
              December 2014, <http://www.rfc-editor.org/info/rfc7435>.

  [I-D.ietf-homenet-prefix-assignment]
              Pfister, P., Paterson, B., and J. Arkko, "Distributed
              Prefix Assignment Algorithm", draft-ietf-homenet-prefix-
              assignment-08 (work in progress), August 2015.

Appendix A.  Alternative Modes of Operation

  Beyond what is described in the main text, the protocol allows for
  other uses.  These are provided as examples.

A.1.  Read-only Operation

  If a node uses just a single endpoint and does not need to publish
  any TLVs, full DNCP node functionality is not required.  Such limited
  node can acquire and maintain view of the TLV space by implementing
  the processing logic as specified in Section 4.4.  Such node would
  not need Trickle, peer-maintenance or even keep-alives at all, as the
  DNCP nodes' use of it would guarantee eventual receipt of network
  state hashes, and synchronization of node data, even in presence of
  unreliable transport.

A.2.  Forwarding Operation

  If a node with a pair of endpoints does not need to publish any TLVs,
  it can detect (for example) nodes with the highest node identifier on
  each of the endpoints (if any).  Any TLVs received from one of them
  would be forwarded verbatim as unicast to the other node with highest
  node identifier.

  Any tinkering with the TLVs would remove guarantees of this scheme
  working; however passive monitoring would obviously be fine.  This
  type of simple forwarding cannot be chained, as it does not send
  anything proactively.

Stenberg & Barth          Expires May 5, 2016                [Page 35]
Internet-Draft    Distributed Node Consensus Protocol    November 2015

Appendix B.  DNCP Profile Additional Guidance

  This appendix explains implications of design choices made when
  specifying DNCP profile to use particular transport or security
  options.

B.1.  Unicast Transport - UDP or TCP?

  The node data published by a DNCP node is limited to 64KB due to the
  16-bit size of the length field of the TLV it is published within.
  Some transport choices may decrease this limit; if using e.g.  UDP
  datagrams for unicast transport the upper bound of node data size is
  whatever the nodes and the underlying network can pass to each other
  as DNCP does not define its own fragmentation scheme.  A profile
  which chooses UDP has to be limited to small node data (e.g. somewhat
  smaller than IPv6 default MTU if using IPv6), or specify a minimum
  which all nodes have to support.  Even then, if using non-link-local
  communications, there is some concern about what middleboxes do to
  fragmented packets.  Therefore, the use of stream transport such as
  TCP is probably a good idea if either non-link-local communication is
  desired, or fragmentation is expected to cause problems.

  TCP also provides some other facilities, such as a relatively long
  built-in keep-alive which in conjunction with connection closes
  occurring from eventual failed retransmissions may be sufficient to
  avoid the use of in-protocol keep-alive defined in Section 6.1.
  Additionally it is reliable, so there is no need for Trickle on such
  unicast connections.

  The major downside of using TCP instead of UDP with DNCP-based
  profiles lies in the loss of control over the time at which TLVs are
  received; while unreliable UDP datagrams also have some delay, TLVs
  within reliable stream transport may be delayed significantly due to
  retransmissions.  This is not a problem if no relative time dependent
  information is stored within the TLVs in the DNCP-based protocol; for
  such a protocol, TCP is a reasonable choice for unicast transport if
  it is available.

B.2.  (Optional) Multicast Transport

  Multicast is needed for dynamic peer discovery and to trigger unicast
  exchanges; for that, unreliable datagram transport (=typically UDP)
  is the only transport option defined within this specification,
  although DNCP-based protocols may themselves define some other
  transport or peer discovery mechanism (e.g. based on mDNS or DNS).

Stenberg & Barth          Expires May 5, 2016                [Page 36]
Internet-Draft    Distributed Node Consensus Protocol    November 2015New version available: draft-baccelli-intarea-adhoc-wireless-com-00.txt