Network Working Group                                            V. Cerf
Request for Comments: 4838              Google/Jet Propulsion Laboratory
Category: Informational                                      S. Burleigh
                                                               A. Hooke
                                                           L. Torgerson
                                         NASA/Jet Propulsion Laboratory
                                                               R. Durst
                                                               K. Scott
                                                  The MITRE Corporation
                                                                K. Fall
                                                      Intel Corporation
                                                               H. Weiss
                                                           SPARTA, Inc.
                                                             April 2007


               Delay-Tolerant Networking Architecture

Status of This Memo

  This memo provides information for the Internet community.  It does
  not specify an Internet standard of any kind.  Distribution of this
  memo is unlimited.

Copyright Notice

  Copyright (C) The IETF Trust (2007).

IESG Note

  This RFC is a product of the Internet Research Task Force and is not
  a candidate for any level of Internet Standard.  The IRTF publishes
  the results of Internet-related research and development activities.
  These results might not be suitable for deployment on the public
  Internet.

Abstract

  This document describes an architecture for delay-tolerant and
  disruption-tolerant networks, and is an evolution of the architecture
  originally designed for the Interplanetary Internet, a communication
  system envisioned to provide Internet-like services across
  interplanetary distances in support of deep space exploration.  This
  document describes an architecture that addresses a variety of
  problems with internetworks having operational and performance
  characteristics that make conventional (Internet-like) networking
  approaches either unworkable or impractical.  We define a message-
  oriented overlay that exists above the transport (or other) layers of



Cerf, et al.                 Informational                      [Page 1]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  the networks it interconnects.  The document presents a motivation
  for the architecture, an architectural overview, review of state
  management required for its operation, and a discussion of
  application design issues.  This document represents the consensus of
  the IRTF DTN research group and has been widely reviewed by that
  group.

Table of Contents

  1. Introduction ....................................................3
  2. Why an Architecture for Delay-Tolerant Networking? ..............4
  3. DTN Architectural Description ...................................5
     3.1. Virtual Message Switching Using Store-and-Forward
          Operation ..................................................5
     3.2. Nodes and Endpoints ........................................7
     3.3. Endpoint Identifiers (EIDs) and Registrations ..............8
     3.4. Anycast and Multicast .....................................10
     3.5. Priority Classes ..........................................10
     3.6. Postal-Style Delivery Options and Administrative Records ..11
     3.7. Primary Bundle Fields .....................................15
     3.8. Routing and Forwarding ....................................16
     3.9. Fragmentation and Reassembly ..............................18
     3.10. Reliability and Custody Transfer .........................19
     3.11. DTN Support for Proxies and Application Layer Gateways ...21
     3.12. Timestamps and Time Synchronization ......................22
     3.13. Congestion and Flow Control at the Bundle Layer ..........22
     3.14. Security .................................................23
  4. State Management Considerations ................................25
     4.1. Application Registration State ............................25
     4.2. Custody Transfer State ....................................26
     4.3. Bundle Routing and Forwarding State .......................26
     4.4. Security-Related State ....................................27
     4.5. Policy and Configuration State ............................27
  5. Application Structuring Issues .................................28
  6. Convergence Layer Considerations for Use of Underlying
     Protocols ......................................................28
  7. Summary ........................................................29
  8. Security Considerations ........................................29
  9. IANA Considerations ............................................30
  10. Normative References ..........................................30
  11. Informative References ........................................30
  12. Acknowledgments ...............................................32









Cerf, et al.                 Informational                      [Page 2]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


1.  Introduction

  This document describes an architecture for delay and disruption-
  tolerant interoperable networking (DTN).  The architecture embraces
  the concepts of occasionally-connected networks that may suffer from
  frequent partitions and that may be comprised of more than one
  divergent set of protocols or protocol families.  The basis for this
  architecture lies with that of the Interplanetary Internet, which
  focused primarily on the issue of deep space communication in high-
  delay environments.  We expect the DTN architecture described here to
  be utilized in various operational environments, including those
  subject to disruption and disconnection and those with high-delay;
  the case of deep space is one specialized example of these, and is
  being pursued as a specialization of this architecture (See [IPN01]
  and [SB03] for more details).

  Other networks to which we believe this architecture applies include
  sensor-based networks using scheduled intermittent connectivity,
  terrestrial wireless networks that cannot ordinarily maintain end-to-
  end connectivity, satellite networks with moderate delays and
  periodic connectivity, and underwater acoustic networks with moderate
  delays and frequent interruptions due to environmental factors.  A
  DTN tutorial [FW03], aimed at introducing DTN and the types of
  networks for which it is designed, is available to introduce new
  readers to the fundamental concepts and motivation.  More technical
  descriptions may be found in [KF03], [JFP04], [JDPF05], and [WJMF05].

  We define an end-to-end message-oriented overlay called the "bundle
  layer" that exists at a layer above the transport (or other) layers
  of the networks on which it is hosted and below applications.
  Devices implementing the bundle layer are called DTN nodes.  The
  bundle layer forms an overlay that employs persistent storage to help
  combat network interruption.  It includes a hop-by-hop transfer of
  reliable delivery responsibility and optional end-to-end
  acknowledgement.  It also includes a number of diagnostic and
  management features.  For interoperability, it uses a flexible naming
  scheme (based on Uniform Resource Identifiers [RFC3986]) capable of
  encapsulating different naming and addressing schemes in the same
  overall naming syntax.  It also has a basic security model,
  optionally enabled, aimed at protecting infrastructure from
  unauthorized use.

  The bundle layer provides functionality similar to the internet layer
  of gateways described in the original ARPANET/Internet designs
  [CK74].  It differs from ARPANET gateways, however, because it is
  layer-agnostic and is focused on virtual message forwarding rather
  than packet switching.  However, both generally provide
  interoperability between underlying protocols specific to one



Cerf, et al.                 Informational                      [Page 3]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  environment and those protocols specific to another, and both provide
  a store-and-forward forwarding service (with the bundle layer
  employing persistent storage for its store and forward function).

  In a sense, the DTN architecture provides a common method for
  interconnecting heterogeneous gateways or proxies that employ store-
  and-forward message routing to overcome communication disruptions.
  It provides services similar to electronic mail, but with enhanced
  naming, routing, and security capabilities.  Nodes unable to support
  the full capabilities required by this architecture may be supported
  by application-layer proxies acting as DTN applications.

2.  Why an Architecture for Delay-Tolerant Networking?

  Our motivation for pursuing an architecture for delay tolerant
  networking stems from several factors.  These factors are summarized
  below; much more detail on their rationale can be explored in [SB03],
  [KF03], and [DFS02].

  The existing Internet protocols do not work well for some
  environments, due to some fundamental assumptions built into the
  Internet architecture:

  - that an end-to-end path between source and destination exists for
    the duration of a communication session

  - (for reliable communication) that retransmissions based on timely
    and stable feedback from data receivers is an effective means for
    repairing errors

  - that end-to-end loss is relatively small

  - that all routers and end stations support the TCP/IP protocols

  - that applications need not worry about communication performance

  - that endpoint-based security mechanisms are sufficient for meeting
    most security concerns

  - that packet switching is the most appropriate abstraction for
    interoperability and performance

  - that selecting a single route between sender and receiver is
    sufficient for achieving acceptable communication performance

  The DTN architecture is conceived to relax most of these assumptions,
  based on a number of design principles that are summarized here (and
  further discussed in [KF03]):



Cerf, et al.                 Informational                      [Page 4]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  - Use variable-length (possibly long) messages (not streams or
    limited-sized packets) as the communication abstraction to help
    enhance the ability of the network to make good scheduling/path
    selection decisions when possible.

  - Use a naming syntax that supports a wide range of naming and
    addressing conventions to enhance interoperability.

  - Use storage within the network to support store-and-forward
    operation over multiple paths, and over potentially long timescales
    (i.e., to support operation in environments where many and/or no
    end-to-end paths may ever exist); do not require end-to-end
    reliability.

  - Provide security mechanisms that protect the infrastructure from
    unauthorized use by discarding traffic as quickly as possible.

  - Provide coarse-grained classes of service, delivery options, and a
    way to express the useful lifetime of data to allow the network to
    better deliver data in serving the needs of applications.

  The use of the bundle layer is guided not only by its own design
  principles, but also by a few application design principles:

  - Applications should minimize the number of round-trip exchanges.

  - Applications should cope with restarts after failure while network
    transactions remain pending.

  - Applications should inform the network of the useful life and
    relative importance of data to be delivered.

  These issues are discussed in further detail in Section 5.

3.  DTN Architectural Description

  The previous section summarized the design principles that guide the
  definition of the DTN architecture.  This section presents a
  description of the major features of the architecture resulting from
  design decisions guided by the aforementioned design principles.

3.1.  Virtual Message Switching Using Store-and-Forward Operation

  A DTN-enabled application sends messages of arbitrary length, also
  called Application Data Units or ADUs [CT90], which are subject to
  any implementation limitations.  The relative order of ADUs might not
  be preserved.  ADUs are typically sent by and delivered to




Cerf, et al.                 Informational                      [Page 5]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  applications in complete units, although a system interface that
  behaves differently is not precluded.

  ADUs are transformed by the bundle layer into one or more protocol
  data units called "bundles", which are forwarded by DTN nodes.
  Bundles have a defined format containing two or more "blocks" of
  data.  Each block may contain either application data or other
  information used to deliver the containing bundle to its
  destination(s).  Blocks serve the purpose of holding information
  typically found in the header or payload portion of protocol data
  units in other protocol architectures.  The term "block" is used
  instead of "header" because blocks may not appear at the beginning of
  a bundle due to particular processing requirements (e.g., digital
  signatures).

  Bundles may be split up ("fragmented") into multiple constituent
  bundles (also called "fragments" or "bundle fragments") during
  transmission.  Fragments are themselves bundles, and may be further
  fragmented.  Two or more fragments may be reassembled anywhere in the
  network, forming a new bundle.

  Bundle sources and destinations are identified by (variable-length)
  Endpoint Identifiers (EIDs, described below), which identify the
  original sender and final destination(s) of bundles, respectively.
  Bundles also contain a "report-to" EID used when special operations
  are requested to direct diagnostic output to an arbitrary entity
  (e.g., other than the source).  An EID may refer to one or more DTN
  nodes (i.e., for multicast destinations or "report-to" destinations).

  While IP networks are based on "store-and-forward" operation, there
  is an assumption that the "storing" will not persist for more than a
  modest amount of time, on the order of the queuing and transmission
  delay.  In contrast, the DTN architecture does not expect that
  network links are always available or reliable, and instead expects
  that nodes may choose to store bundles for some time.  We anticipate
  that most DTN nodes will use some form of persistent storage for this
  -- disk, flash memory, etc. -- and that stored bundles will survive
  system restarts.

  Bundles contain an originating timestamp, useful life indicator, a
  class of service designator, and a length.  This information provides
  bundle-layer routing with a priori knowledge of the size and
  performance requirements of requested data transfers.  When there is
  a significant amount of queuing that can occur in the network (as is
  the case in the DTN version of store-and-forward), the advantage
  provided by knowing this information may be significant for making
  scheduling and path selection decisions [JFP04].  An alternative
  abstraction (i.e., of stream-based delivery based on packets) would



Cerf, et al.                 Informational                      [Page 6]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  make such scheduling much more difficult.  Although packets provide
  some of the same benefits as bundles, larger aggregates provide a way
  for the network to apply scheduling and buffer management to units of
  data that are more useful to applications.

  An essential element of the bundle-based style of forwarding is that
  bundles have a place to wait in a queue until a communication
  opportunity ("contact") is available.  This highlights the following
  assumptions:

  1. that storage is available and well-distributed throughout the
     network,

  2. that storage is sufficiently persistent and robust to store
     bundles until forwarding can occur, and

  3. (implicitly) that this "store-and-forward" model is a better
     choice than attempting to effect continuous connectivity or other
     alternatives.

  For a network to effectively support the DTN architecture, these
  assumptions must be considered and must be found to hold.  Even so,
  the inclusion of long-term storage as a fundamental aspect of the DTN
  architecture poses new problems, especially with respect to
  congestion management and denial-of-service mitigation.  Node storage
  in essence represents a new resource that must be managed and
  protected.  Much of the research in DTN revolves around exploring
  these issues.  Congestion is discussed in Section 3.13, and security
  mechanisms, including methods for DTN nodes to protect themselves
  from handling unauthorized traffic from other nodes, are discussed in
  [DTNSEC] and [DTNSOV].

3.2.  Nodes and Endpoints

  A DTN node (or simply "node" in this document) is an engine for
  sending and receiving bundles -- an implementation of the bundle
  layer.  Applications utilize DTN nodes to send or receive ADUs
  carried in bundles (applications also use DTN nodes when acting as
  report-to destinations for diagnostic information carried in
  bundles).  Nodes may be members of groups called "DTN endpoints".  A
  DTN endpoint is therefore a set of DTN nodes.  A bundle is considered
  to have been successfully delivered to a DTN endpoint when some
  minimum subset of the nodes in the endpoint has received the bundle
  without error.  This subset is called the "minimum reception group"
  (MRG) of the endpoint.  The MRG of an endpoint may refer to one node
  (unicast), one of a group of nodes (anycast), or all of a group of
  nodes (multicast and broadcast).  A single node may be in the MRG of
  multiple endpoints.



Cerf, et al.                 Informational                      [Page 7]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


3.3.  Endpoint Identifiers (EIDs) and Registrations

  An Endpoint Identifier (EID) is a name, expressed using the general
  syntax of URIs (see below), that identifies a DTN endpoint.  Using an
  EID, a node is able to determine the MRG of the DTN endpoint named by
  the EID.  Each node is also required to have at least one EID that
  uniquely identifies it.

  Applications send ADUs destined for an EID, and may arrange for ADUs
  sent to a particular EID to be delivered to them.  Depending on the
  construction of the EID being used (see below), there may be a
  provision for wildcarding some portion of an EID, which is often
  useful for diagnostic and routing purposes.

  An application's desire to receive ADUs destined for a particular EID
  is called a "registration", and in general is maintained persistently
  by a DTN node.  This allows application registration information to
  survive application and operating system restarts.

  An application's attempt to establish a registration is not
  guaranteed to succeed.  For example, an application could request to
  register itself to receive ADUs by specifying an Endpoint ID that is
  uninterpretable or unavailable to the DTN node servicing the request.
  Such requests are likely to fail.

3.3.1.  URI Schemes

  Each Endpoint ID is expressed syntactically as a Uniform Resource
  Identifier (URI) [RFC3986].  The URI syntax has been designed as a
  way to express names or addresses for a wide range of purposes, and
  is therefore useful for constructing names for DTN endpoints.

  In URI terminology, each URI begins with a scheme name.  The scheme
  name is an element of the set of globally-managed scheme names
  maintained by IANA [ISCHEMES].  Lexically following the scheme name
  in a URI is a series of characters constrained by the syntax defined
  by the scheme.  This portion of the URI is called the scheme-specific
  part (SSP), and can be quite general.  (See, as one example, the URI
  scheme for SNMP [RFC4088]).  Note that scheme-specific syntactical
  and semantic restrictions may be more constraining than the basic
  rules of RFC 3986.  Section 3.1 of RFC 3986 provides guidance on the
  syntax of scheme names.

  URI schemes are a key concept in the DTN architecture, and evolved
  from an earlier concept called regions, which were tied more closely
  to assumptions of the network topology.  Using URIs, significant
  flexibility is attained in the structuring of EIDs.  They might, for
  example, be constructed based on DNS names, or might look like



Cerf, et al.                 Informational                      [Page 8]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  "expressions of interest" or forms of database-like queries as in a
  directed diffusion-routed network [IGE00] or in intentional naming
  [WSBL99].  As names, EIDs are not required to be related to routing
  or topological organization.  Such a relationship is not prohibited,
  however, and in some environments using EIDs this way may be
  advantageous.

  A single EID may refer to an endpoint containing more than one DTN
  node, as suggested above.  It is the responsibility of a scheme
  designer to define how to interpret the SSP of an EID so as to
  determine whether it refers to a unicast, multicast, or anycast set
  of nodes.  See Section 3.4 for more details.

  URIs are constructed based on rules specified in RFC 3986, using the
  US-ASCII character set.  However, note this excerpt from RFC 3986,
  Section 1.2.1, on dealing with characters that cannot be represented
  by US-ASCII:  "Percent-encoded octets (Section 2.1) may be used
  within a URI to represent characters outside the range of the US-
  ASCII coded character set if this representation is allowed by the
  scheme or by the protocol element in which the URI is referenced.
  Such a definition should specify the character encoding used to map
  those characters to octets prior to being percent-encoded for the
  URI".

3.3.2.  Late Binding

  Binding means interpreting the SSP of an EID for the purpose of
  carrying an associated message towards a destination.  For example,
  binding might require mapping an EID to a next-hop EID or to a lower-
  layer address for transmission.  "Late binding" means that the
  binding of a bundle's destination to a particular set of destination
  identifiers or addresses does not necessarily happen at the bundle
  source.  Because the destination EID is potentially re-interpreted at
  each hop, the binding may occur at the source, during transit, or
  possibly at the destination(s).  This contrasts with the name-to-
  address binding of Internet communications where a DNS lookup at the
  source fixes the IP address of the destination node before data is
  sent.  Such a circumstance would be considered "early binding"
  because the name-to-address translation is performed prior to data
  being sent into the network.

  In a frequently-disconnected network, late binding may be
  advantageous because the transit time of a message may exceed the
  validity time of a binding, making binding at the source impossible
  or invalid.  Furthermore, use of name-based routing with late binding
  may reduce the amount of administrative (mapping) information that





Cerf, et al.                 Informational                      [Page 9]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  must propagate through the network, and may also limit the scope of
  mapping synchronization requirements to a local topological
  neighborhood where changes are made.

3.4.  Anycast and Multicast

  As mentioned above, an EID may refer to an endpoint containing one or
  more DTN nodes.  When referring to a group of size greater than one,
  the delivery semantics may be of either the anycast or multicast
  variety (broadcast is considered to be of the multicast variety).
  For anycast group delivery, a bundle is delivered to one node among a
  group of potentially many nodes, and for multicast delivery it is
  intended to be delivered to all of them, subject to the normal DTN
  class of service and maximum useful lifetime semantics.

  Multicast group delivery in a DTN presents an unfamiliar issue with
  respect to group membership.  In relatively low-delay networks, such
  as the Internet, nodes may be considered to be part of the group if
  they have expressed interest to join it "recently".  In a DTN,
  however, nodes may wish to receive data sent to a group during an
  interval of time earlier than when they are actually able to receive
  it [ZAZ05].  More precisely, an application expresses its desire to
  receive data sent to EID e at time t.  Prior to this, during the
  interval [t0, t1], t > t1, data may have been generated for group e.
  For the application to receive any of this data, the data must be
  available a potentially long time after senders have ceased sending
  to the group.  Thus, the data may need to be stored within the
  network in order to support temporal group semantics of this kind.
  How to design and implement this remains a research issue, as it is
  likely to be at least as hard as problems related to reliable
  multicast.

3.5.  Priority Classes

  The DTN architecture offers *relative* measures of priority (low,
  medium, high) for delivering ADUs.  These priorities differentiate
  traffic based upon an application's desire to affect the delivery
  urgency for ADUs, and are carried in bundle blocks generated by the
  bundle layer based on information specified by the application.

  The (U.S. or similar) Postal Service provides a strong metaphor for
  the priority classes offered by the forwarding abstraction offered by
  the DTN architecture.  Traffic is generally not interactive and is
  often one-way.  There are generally no strong guarantees of timely
  delivery, yet there are some forms of class of service, reliability,
  and security.





Cerf, et al.                 Informational                     [Page 10]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  We have defined three relative priority classes to date.  These
  priority classes typically imply some relative scheduling
  prioritization among bundles in queue at a sender:

  - Bulk - Bulk bundles are shipped on a "least effort" basis.  No
    bundles of this class will be shipped until all bundles of other
    classes bound for the same destination and originating from the
    same source have been shipped.

  - Normal - Normal-class bundles are shipped prior to any bulk-class
    bundles and are otherwise the same as bulk bundles.

  - Expedited - Expedited bundles, in general, are shipped prior to
    bundles of other classes and are otherwise the same.

  Applications specify their requested priority class and data lifetime
  (see below) for each ADU they send.  This information, coupled with
  policy applied at DTN nodes that select how messages are forwarded
  and which routing algorithms are in use, affects the overall
  likelihood and timeliness of ADU delivery.

  The priority class of a bundle is only required to relate to other
  bundles from the same source.  This means that a high priority bundle
  from one source may not be delivered faster (or with some other
  superior quality of service) than a medium priority bundle from a
  different source.  It does mean that a high priority bundle from one
  source will be handled preferentially to a lower priority bundle sent
  from the same source.

  Depending on a particular DTN node's forwarding/scheduling policy,
  priority may or may not be enforced across different sources.  That
  is, in some DTN nodes, expedited bundles might always be sent prior
  to any bulk bundles, irrespective of source.  Many variations are
  possible.

3.6.  Postal-Style Delivery Options and Administrative Records

  Continuing with the postal analogy, the DTN architecture supports
  several delivery options that may be selected by an application when
  it requests the transmission of an ADU.  In addition, the
  architecture defines two types of administrative records: "status
  reports" and "signals".  These records are bundles that provide
  information about the delivery of other bundles, and are used in
  conjunction with the delivery options.







Cerf, et al.                 Informational                     [Page 11]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


3.6.1.  Delivery Options

  We have defined eight delivery options.  Applications sending an ADU
  (the "subject ADU") may request any combination of the following,
  which are carried in each of the bundles produced ("sent bundles") by
  the bundle layer resulting from the application's request to send the
  subject ADU:

  - Custody Transfer Requested - requests sent bundles be delivered
    with enhanced reliability using custody transfer procedures.  Sent
    bundles will be transmitted by the bundle layer using reliable
    transfer protocols (if available), and the responsibility for
    reliable delivery of the bundle to its destination(s) may move
    among one or more "custodians" in the network.  This capability is
    described in more detail in Section 3.10.

  - Source Node Custody Acceptance Required - requires the source DTN
    node to provide custody transfer for the sent bundles.  If custody
    transfer is not available at the source when this delivery option
    is requested, the requested transmission fails.  This provides a
    means for applications to insist that the source DTN node take
    custody of the sent bundles (e.g., by storing them in persistent
    storage).

  - Report When Bundle Delivered - requests a (single) Bundle Delivery
    Status Report be generated when the subject ADU is delivered to its
    intended recipient(s).  This request is also known as "return-
    receipt".

  - Report When Bundle Acknowledged by Application - requests an
    Acknowledgement Status Report be generated when the subject ADU is
    acknowledged by a receiving application.  This only happens by
    action of the receiving application, and differs from the Bundle
    Delivery Status Report.  It is intended for cases where the
    application may be acting as a form of application layer gateway
    and wishes to indicate the status of a protocol operation external
    to DTN back to the requesting source.  See Section 11 for more
    details.

  - Report When Bundle Received - requests a Bundle Reception Status
    Report be generated when each sent bundle arrives at a DTN node.
    This is designed primarily for diagnostic purposes.

  - Report When Bundle Custody Accepted  - requests a Custody
    Acceptance Status Report be generated when each sent bundle has
    been accepted using custody transfer.  This is designed primarily
    for diagnostic purposes.




Cerf, et al.                 Informational                     [Page 12]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  - Report When Bundle Forwarded - requests a Bundle Forwarding Status
    Report be generated when each sent bundle departs a DTN node after
    forwarding.  This is designed primarily for diagnostic purposes.

  - Report When Bundle Deleted - requests a Bundle Deletion Status
    Report be generated when each sent bundle is deleted at a DTN node.
    This is designed primarily for diagnostic purposes.

  The first four delivery options are designed for ordinary use by
  applications.  The last four are designed primarily for diagnostic
  purposes and their use may be restricted or limited in environments
  subject to congestion or attack.

  If the security procedures defined in [DTNSEC] are also enabled, then
  three additional delivery options become available:

  - Confidentiality Required - requires the subject ADU be made secret
    from parties other than the source and the members of the
    destination EID.

  - Authentication Required - requires all non-mutable fields in the
    bundle blocks of the sent bundles (i.e., those which do not change
    as the bundle is forwarded) be made strongly verifiable (i.e.,
    cryptographically strong).  This protects several fields, including
    the source and destination EIDs and the bundle's data.  See Section
    3.7 and [BSPEC] for more details.

  - Error Detection Required - requires modifications to the non-
    mutable fields of each sent bundle be made detectable with high
    probability at each destination.

3.6.2.  Administrative Records: Bundle Status Reports and Custody
       Signals

  Administrative records are used to report status information or error
  conditions related to the bundle layer.  There are two types of
  administrative records defined:  bundle status reports (BSRs) and
  custody signals.  Administrative records correspond (approximately)
  to messages in the ICMP protocol in IP [RFC792].  In ICMP, however,
  messages are returned to the source.  In DTN, they are instead
  directed to the report-to EID for BSRs and the EID of the current
  custodian for custody signals, which might differ from the source's
  EID.  Administrative records are sent as bundles with a source EID
  set to one of the EIDs associated with the DTN node generating the
  administrative record.  In some cases, arrival of a single bundle or
  bundle fragment may elicit multiple administrative records (e.g., in
  the case where a bundle is replicated for multicast forwarding).




Cerf, et al.                 Informational                     [Page 13]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  The following BSRs are currently defined (also see [BSPEC] for more
  details):

  - Bundle Reception - sent when a bundle arrives at a DTN node.
    Generation of this message may be limited by local policy.

  - Custody Acceptance - sent when a node has accepted custody of a
    bundle with the Custody Transfer Requested option set.  Generation
    of this message may be limited by local policy.

  - Bundle Forwarded - sent when a bundle containing a Report When
    Bundle Forwarded option departs from a DTN node after having been
    forwarded.  Generation of this message may be limited by local
    policy.

  - Bundle Deletion - sent from a DTN node when a bundle containing a
    Report When Bundle Deleted option is discarded.  This can happen
    for several reasons, such as expiration.  Generation of this
    message may be limited by local policy but is required in cases
    where the deletion is performed by a bundle's current custodian.

  - Bundle Delivery - sent from a final recipient's (destination) node
    when a complete ADU comprising sent bundles containing Report When
    Bundle Delivered options is consumed by an application.

  - Acknowledged by application - sent from a final recipient's
    (destination) node when a complete ADU comprising sent bundles
    containing Application Acknowledgment options has been processed by
    an application.  This generally involves specific action on the
    receiving application's part.

  In addition to the status reports, the custody signal is currently
  defined to indicate the status of a custody transfer.  These are sent
  to the current-custodian EID contained in an arriving bundle:

  - Custody Signal - indicates that custody has been successfully
    transferred.  This signal appears as a Boolean indicator, and may
    therefore indicate either a successful or a failed custody transfer
    attempt.

  Administrative records must reference a received bundle.  This is
  accomplished by a method for uniquely identifying bundles based on a
  transmission timestamp and sequence number discussed in Section 3.12.








Cerf, et al.                 Informational                     [Page 14]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


3.7.  Primary Bundle Fields

  The bundles carried between and among DTN nodes obey a standard
  bundle protocol specified in [BSPEC].  Here we provide an overview of
  most of the fields carried with every bundle.  The protocol is
  designed with a mandatory primary block, an optional payload block
  (which contains the ADU data itself), and a set of optional extension
  blocks.  Blocks may be cascaded in a way similar to extension headers
  in IPv6.  The following selected fields are all present in the
  primary block, and therefore are present for every bundle and
  fragment:

  - Creation Timestamp - a concatenation of the bundle's creation time
    and a monotonically increasing sequence number such that the
    creation timestamp is guaranteed to be unique for each ADU
    originating from the same source.  The creation timestamp is based
    on the time-of-day an application requested an ADU to be sent (not
    when the corresponding bundle(s) are sent into the network).  DTN
    nodes are assumed to have a basic time synchronization capability
    (see Section 3.12).

  - Lifespan - the time-of-day at which the message is no longer
    useful.  If a bundle is stored in the network (including the
    source's DTN node) when its lifespan is reached, it may be
    discarded.  The lifespan of a bundle is expressed as an offset
    relative to its creation time.

  - Class of Service Flags - indicates the delivery options and
    priority class for the bundle.  Priority classes may be one of
    bulk, normal, or expedited.  See Section 3.6.1.

  - Source EID - EID of the source (the first sender).

  - Destination EID - EID of the destination (the final intended
    recipient(s)).

  - Report-To Endpoint ID - an EID identifying where reports (return-
    receipt, route-tracing functions) should be sent.  This may or may
    not identify the same endpoint as the Source EID.

  - Custodian EID - EID of the current custodian of a bundle (if any).

  The payload block indicates information about the contained payload
  (e.g., its length) and the payload itself.  In addition to the fields
  found in the primary and payload blocks, each bundle may have fields
  in additional blocks carried with each bundle.  See [BSPEC] for
  additional details.




Cerf, et al.                 Informational                     [Page 15]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


3.8.  Routing and Forwarding

  The DTN architecture provides a framework for routing and forwarding
  at the bundle layer for unicast, anycast, and multicast messages.
  Because nodes in a DTN network might be interconnected using more
  than one type of underlying network technology, a DTN network is best
  described abstractly using a *multigraph* (a graph where vertices may
  be interconnected with more than one edge).  Edges in this graph are,
  in general, time-varying with respect to their delay and capacity and
  directional because of the possibility of one-way connectivity.  When
  an edge has zero capacity, it is considered to not be connected.

  Because edges in a DTN graph may have significant delay, it is
  important to distinguish where time is measured when expressing an
  edge's capacity or delay.  We adopt the convention of expressing
  capacity and delay as functions of time where time is measured at the
  point where data is inserted into a network edge.  For example,
  consider an edge having capacity C(t) and delay D(t) at time t.  If B
  bits are placed in this edge at time t, they completely arrive by
  time t + D(t) + (1/C(t))*B.  We assume C(t) and D(t) do not change
  significantly during the interval [t, t+D(t)+(1/C(t))*B].

  Because edges may vary between positive and zero capacity, it is
  possible to describe a period of time (interval) during which the
  capacity is strictly positive, and the delay and capacity can be
  considered to be constant [AF03].  This period of time is called a
  "contact".  In addition, the product of the capacity and the interval
  is known as a contact's "volume".  If contacts and their volumes are
  known ahead of time, intelligent routing and forwarding decisions can
  be made (optimally for small networks) [JFP04].  Optimally using a
  contact's volume, however, requires the ability to divide large ADUs
  and bundles into smaller routable units.  This is provided by DTN
  fragmentation (see Section 3.9).

  When delivery paths through a DTN graph are lossy or contact
  intervals and volumes are not known precisely ahead of time, routing
  computations become especially challenging.  How to handle these
  situations is an active area of work in the (emerging) research area
  of delay tolerant networking.

3.8.1.  Types of Contacts

  Contacts typically fall into one of several categories, based largely
  on the predictability of their performance characteristics and
  whether some action is required to bring them into existence.  To
  date, the following major types of contacts have been defined:





Cerf, et al.                 Informational                     [Page 16]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  Persistent Contacts

     Persistent contacts are always available (i.e., no connection-
     initiation action is required to instantiate a persistent
     contact).  An 'always-on' Internet connection such as a DSL or
     Cable Modem connection would be a representative of this class.

  On-Demand Contacts

     On-Demand contacts require some action in order to instantiate,
     but then function as persistent contacts until terminated.  A
     dial-up connection is an example of an On-Demand contact (at
     least, from the viewpoint of the dialer; it may be viewed as an
     Opportunistic Contact, below, from the viewpoint of the dial-up
     service provider).

  Intermittent - Scheduled Contacts

     A scheduled contact is an agreement to establish a contact at a
     particular time, for a particular duration.  An example of a
     scheduled contact is a link with a low-earth orbiting satellite.
     A node's list of contacts with the satellite can be constructed
     from the satellite's schedule of view times, capacities, and
     latencies.  Note that for networks with substantial delays, the
     notion of the "particular time" is delay-dependent.  For example,
     a single scheduled contact between Earth and Mars would not be at
     the same instant in each location, but would instead be offset by
     the (non-negligible) propagation delay.

  Intermittent - Opportunistic Contacts

     Opportunistic contacts are not scheduled, but rather present
     themselves unexpectedly.  For example, an unscheduled aircraft
     flying overhead and beaconing, advertising its availability for
     communication, would present an opportunistic contact.  Another
     type of opportunistic contact might be via an infrared or
     Bluetooth communication link between a personal digital assistant
     (PDA) and a kiosk in an airport concourse.  The opportunistic
     contact begins as the PDA is brought near the kiosk, lasting an
     undetermined amount of time (i.e., until the link is lost or
     terminated).

  Intermittent - Predicted Contacts

     Predicted contacts are based on no fixed schedule, but rather are
     predictions of likely contact times and durations based on a
     history of previously observed contacts or some other information.
     Given a great enough confidence in a predicted contact, routes may



Cerf, et al.                 Informational                     [Page 17]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


     be chosen based on this information.  This is an active research
     area, and a few approaches having been proposed [LFC05].

3.9.  Fragmentation and Reassembly

  DTN fragmentation and reassembly are designed to improve the
  efficiency of bundle transfers by ensuring that contact volumes are
  fully utilized and by avoiding retransmission of partially-forwarded
  bundles.  There are two forms of DTN fragmentation/reassembly:

  Proactive Fragmentation

     A DTN node may divide a block of application data into multiple
     smaller blocks and transmit each such block as an independent
     bundle.  In this case, the *final destination(s)* are responsible
     for extracting the smaller blocks from incoming bundles and
     reassembling them into the original larger bundle and, ultimately,
     ADU.  This approach is called proactive fragmentation because it
     is used primarily when contact volumes are known (or predicted) in
     advance.

  Reactive Fragmentation

     DTN nodes sharing an edge in the DTN graph may fragment a bundle
     cooperatively when a bundle is only partially transferred.  In
     this case, the receiving bundle layer modifies the incoming bundle
     to indicate it is a fragment, and forwards it normally.  The
     previous- hop sender may learn (via convergence-layer protocols,
     see Section 6) that only a portion of the bundle was delivered to
     the next hop, and send the remaining portion(s) when subsequent
     contacts become available (possibly to different next-hops if
     routing changes).  This is called reactive fragmentation because
     the fragmentation process occurs after an attempted transmission
     has taken place.

     As an example, consider a ground station G, and two store-and-
     forward satellites S1 and S2, in opposite low-earth orbit.  While
     G is transmitting a large bundle to S1, a reliable transport layer
     protocol below the bundle layer at each indicates the transmission
     has terminated, but that half the transfer has completed
     successfully.  In this case, G can form a smaller bundle fragment
     consisting of the second half of the original bundle and forward
     it to S2 when available.  In addition, S1 (now out of range of G)
     can form a new bundle consisting of the first half of the original
     bundle and forward it to whatever next hop(s) it deems
     appropriate.





Cerf, et al.                 Informational                     [Page 18]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  The reactive fragmentation capability is not required to be available
  in every DTN implementation, as it requires a certain level of
  support from underlying protocols that may not be present, and
  presents significant challenges with respect to handling digital
  signatures and authentication codes on messages.  When a signed
  message is only partially received, most message authentication codes
  will fail.  When DTN security is present and enabled, it may
  therefore be necessary to proactively fragment large bundles into
  smaller units that are more convenient for digital signatures.

  Even if reactive fragmentation is not present in an implementation,
  the ability to reassemble fragments at a destination is required in
  order to support DTN fragmentation.  Furthermore, for contacts with
  volumes that are small compared to typical bundle sizes, some
  incremental delivery approach must be used (e.g., checkpoint/restart)
  to prevent data delivery livelock.  Reactive fragmentation is one
  such approach, but other protocol layers could potentially handle
  this issue as well.

3.10.  Reliability and Custody Transfer

  The most basic service provided by the bundle layer is
  unacknowledged, prioritized (but not guaranteed) unicast message
  delivery.  It also provides two options for enhancing delivery
  reliability:  end-to-end acknowledgments and custody transfer.
  Applications wishing to implement their own end-to-end message
  reliability mechanisms are free to utilize the acknowledgment.  The
  custody transfer feature of the DTN architecture only specifies a
  coarse-grained retransmission capability, described next.

  Transmission of bundles with the Custody Transfer Requested option
  specified generally involves moving the responsibility for reliable
  delivery of an ADU's bundles among different DTN nodes in the
  network.  For unicast delivery, this will typically involve moving
  bundles "closer" (in terms of some routing metric) to their ultimate
  destination(s), and retransmitting when necessary.  The nodes
  receiving these bundles along the way (and agreeing to accept the
  reliable delivery responsibility) are called "custodians".  The
  movement of a bundle (and its delivery responsibility) from one node
  to another is called a "custody transfer".  It is analogous to a
  database commit transaction [FHM03].  The exact meaning and design of
  custody transfer for multicast and anycast delivery remains to be
  fully explored.

  Custody transfer allows the source to delegate retransmission
  responsibility and recover its retransmission-related resources
  relatively soon after sending a bundle (on the order of the minimum
  round-trip time to the first bundle hop(s)).  Not all nodes in a DTN



Cerf, et al.                 Informational                     [Page 19]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  are required by the DTN architecture to accept custody transfers, so
  it is not a true 'hop-by-hop' mechanism.  For example, some nodes may
  have sufficient storage resources to sometimes act as custodians, but
  may elect to not offer such services when congested or running low on
  power.

  The existence of custodians can alter the way DTN routing is
  performed.  In some circumstances, it may be beneficial to move a
  bundle to a custodian as quickly as possible even if the custodian is
  further away (in terms of distance, time or some routing metric) from
  the bundle's final destination(s) than some other reachable node.
  Designing a system with this capability involves constructing more
  than one routing graph, and is an area of continued research.

  Custody transfer in DTN not only provides a method for tracking
  bundles that require special handling and identifying DTN nodes that
  participate in custody transfer, it also provides a (weak) mechanism
  for enhancing the reliability of message delivery.  Generally
  speaking, custody transfer relies on underlying reliable delivery
  protocols of the networks that it operates over to provide the
  primary means of reliable transfer from one bundle node to the next
  (set).  However, when custody transfer is requested, the bundle layer
  provides an additional coarse-grained timeout and retransmission
  mechanism and an accompanying (bundle-layer) custodian-to-custodian
  acknowledgment signaling mechanism.  When an application does *not*
  request custody transfer, this bundle layer timeout and
  retransmission mechanism is typically not employed, and successful
  bundle layer delivery depends solely on the reliability mechanisms of
  the underlying protocols.

  When a node accepts custody for a bundle that contains the Custody
  Transfer Requested option, a Custody Transfer Accepted Signal is sent
  by the bundle layer to the Current Custodian EID contained in the
  primary bundle block.  In addition, the Current Custodian EID is
  updated to contain one of the forwarding node's (unicast) EIDs before
  the bundle is forwarded.

  When an application requests an ADU to be delivered with custody
  transfer, the request is advisory.  In some circumstances, a source
  of a bundle for which custody transfer has been requested may not be
  able to provide this service.  In such circumstances, the subject
  bundle may traverse multiple DTN nodes before it obtains a custodian.
  Bundles in this condition are specially marked with their Current
  Custodian EID field set to a null endpoint.  In cases where
  applications wish to require the source to take custody of the
  bundle, they may supply the Source Node Custody Acceptance Required





Cerf, et al.                 Informational                     [Page 20]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  delivery option.  This may be useful to applications that desire a
  continuous "chain" of custody or that wish to exit after being
  ensured their data is safely held in a custodian.

  In a DTN network where one or more custodian-to-custodian hops are
  strictly one directional (and cannot be reversed), the DTN custody
  transfer mechanism will be affected over such hops due to the lack of
  any way to receive a custody signal (or any other information) back
  across the path, resulting in the expiration of the bundle at the
  ingress to the one-way hop.  This situation does not necessarily mean
  the bundle has been lost; nodes on the other side of the hop may
  continue to transfer custody, and the bundle may be delivered
  successfully to its destination(s).  However, in this circumstance a
  source that has requested to receive expiration BSRs for this bundle
  will receive an expiration report for the bundle, and possibly
  conclude (incorrectly) that the bundle has been discarded and not
  delivered.  Although this problem cannot be fully solved in this
  situation, a mechanism is provided to help ameliorate the seemingly
  incorrect information that may be reported when the bundle expires
  after having been transferred over a one-way hop.  This is
  accomplished by the node at the ingress to the one-way hop reporting
  the existence of a known one-way path using a variant of a bundle
  status report.  These types of reports are provided if the subject
  bundle requests the report using the 'Report When Bundle Forwarded'
  delivery option.

3.11.  DTN Support for Proxies and Application Layer Gateways

  One of the aims of DTN is to provide a common method for
  interconnecting application layer gateways and proxies.  In cases
  where existing Internet applications can be made to tolerate delays,
  local proxies can be constructed to benefit from the existing
  communication capabilities provided by DTN [S05, T02].  Making such
  proxies compatible with DTN reduces the burden on the proxy author
  from being concerned with how to implement routing and reliability
  management and allows existing TCP/IP-based applications to operate
  unmodified over a DTN-based network.

  When DTN is used to provide a form of tunnel encapsulation for other
  protocols, it can be used in constructing overlay networks comprised
  of application layer gateways.  The application acknowledgment
  capability is designed for such circumstances.  This provides a
  common way for remote application layer gateways to signal the
  success or failure of non-DTN protocol operations initiated as a
  result of receiving DTN ADUs.  Without this capability, such
  indicators would have to be implemented by applications themselves in
  non-standard ways.




Cerf, et al.                 Informational                     [Page 21]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


3.12.  Timestamps and Time Synchronization

  The DTN architecture depends on time synchronization among DTN nodes
  (supported by external, non-DTN protocols) for four primary purposes:
  bundle and fragment identification, routing with scheduled or
  predicted contacts, bundle expiration time computations, and
  application registration expiration.

  Bundle identification and expiration are supported by placing a
  creation timestamp and an explicit expiration field (expressed in
  seconds after the source timestamp) in each bundle.  The origination
  timestamps on arriving bundles are made available to consuming
  applications in ADUs they receive by some system interface function.
  Each set of bundles corresponding to an ADU is required to contain a
  timestamp unique to the sender's EID.  The EID, timestamp, and data
  offset/length information together uniquely identify a bundle.
  Unique bundle identification is used for a number of purposes,
  including custody transfer and reassembly of bundle fragments.

  Time is also used in conjunction with application registrations.
  When an application expresses its desire to receive ADUs destined for
  a particular EID, this registration is only maintained for a finite
  period of time, and may be specified by the application.  For
  multicast registrations, an application may also specify a time range
  or "interest interval" for its registration.  In this case, traffic
  sent to the specified EID any time during the specified interval will
  eventually be delivered to the application (unless such traffic has
  expired due to the expiration time provided by the application at the
  source or some other reason prevents such delivery).

3.13.  Congestion and Flow Control at the Bundle Layer

  The subject of congestion control and flow control at the bundle
  layer is one on which the authors of this document have not yet
  reached complete consensus.  We have unresolved concerns about the
  efficiency and efficacy of congestion and flow control schemes
  implemented across long and/or highly variable delay environments,
  especially with the custody transfer mechanism that may require nodes
  to retain bundles for long periods of time.

  For the purposes of this document, we define "flow control" as a
  means of assuring that the average rate at which a sending node
  transmits data to a receiving node does not exceed the average rate
  at which the receiving node is prepared to receive data from that
  sender. (Note that this is a generalized notion of flow control,
  rather than one that applies only to end-to-end communication.)  We
  define "congestion control" as a means of assuring that the aggregate
  rate at which all traffic sources inject data into a network does not



Cerf, et al.                 Informational                     [Page 22]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  exceed the maximum aggregate rate at which the network can deliver
  data to destination nodes over time.  If flow control is propagated
  backward from congested nodes toward traffic sources, then the flow
  control mechanism can be used as at least a partial solution to the
  problem of congestion as well.

  DTN flow control decisions must be made within the bundle layer
  itself based on information about resources (in this case, primarily
  persistent storage) available within the bundle node.  When storage
  resources become scarce, a DTN node has only a certain degree of
  freedom in handling the situation.  It can always discard bundles
  which have expired -- an activity DTN nodes should perform regularly
  in any case.  If it ordinarily is willing to accept custody for
  bundles, it can cease doing so.  If storage resources are available
  elsewhere in the network, it may be able to make use of them in some
  way for bundle storage.  It can also discard bundles which have not
  expired but for which it has not accepted custody.  A node must avoid
  discarding bundles for which it has accepted custody, and do so only
  as a last resort.  Determining when a node should engage in or cease
  to engage in custody transfers is a resource allocation and
  scheduling problem of current research interest.

  In addition to the bundle layer mechanisms described above, a DTN
  node may be able to avail itself of support from lower-layer
  protocols in affecting its own resource utilization.  For example, a
  DTN node receiving a bundle using TCP/IP might intentionally slow
  down its receiving rate by performing read operations less frequently
  in order to reduce its offered load.  This is possible because TCP
  provides its own flow control, so reducing the application data
  consumption rate could effectively implement a form of hop-by-hop
  flow control.  Unfortunately, it may also lead to head-of-line
  blocking issues, depending on the nature of bundle multiplexing
  within a TCP connection.  A protocol with more relaxed ordering
  constraints (e.g. SCTP [RFC2960]) might be preferable in such
  circumstances.

  Congestion control is an ongoing research topic.

3.14.  Security

  The possibility of severe resource scarcity in some delay-tolerant
  networks dictates that some form of authentication and access control
  to the network itself is required in many circumstances.  It is not
  acceptable for an unauthorized user to flood the network with traffic
  easily, possibly denying service to authorized users.  In many cases
  it is also not acceptable for unauthorized traffic to be forwarded
  over certain network links at all.  This is especially true for
  exotic, mission-critical links.  In light of these considerations,



Cerf, et al.                 Informational                     [Page 23]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  several goals are established for the security component of the DTN
  architecture:

  - Promptly prevent unauthorized applications from having their data
    carried through or stored in the DTN.

  - Prevent unauthorized applications from asserting control over the
    DTN infrastructure.

  - Prevent otherwise authorized applications from sending bundles at a
    rate or class of service for which they lack permission.

  - Promptly discard bundles that are damaged or improperly modified in
    transit.

  - Promptly detect and de-authorize compromised entities.

  Many existing authentication and access control protocols designed
  for operation in low-delay, connected environments may not perform
  well in DTNs.  In particular, updating access control lists and
  revoking ("blacklisting") credentials may be especially difficult.
  Also, approaches that require frequent access to centralized servers
  to complete an authentication or authorization transaction are not
  attractive.  The consequences of these difficulties include delays in
  the onset of communication, delays in detecting and recovering from
  system compromise, and delays in completing transactions due to
  inappropriate access control or authentication settings.

  To help satisfy these security requirements in light of the
  challenges, the DTN architecture adopts a standard but optionally
  deployed security architecture [DTNSEC] that utilizes hop-by-hop and
  end-to-end authentication and integrity mechanisms.  The purpose of
  using both approaches is to be able to handle access control for data
  forwarding and storage separately from application-layer data
  integrity.  While the end-to-end mechanism provides authentication
  for a principal such as a user (of which there may be many), the hop-
  by-hop mechanism is intended to authenticate DTN nodes as legitimate
  transceivers of bundles to each-other.  Note that it is conceivable
  to construct a DTN in which only a subset of the nodes participate in
  the security mechanisms, resulting in a secure DTN overlay existing
  atop an insecure DTN overlay.  This idea is relatively new and is
  still being explored.

  In accordance with the goals listed above, DTN nodes discard traffic
  as early as possible if authentication or access control checks fail.
  This approach meets the goals of removing unwanted traffic from being
  forwarded over specific high-value links, but also has the associated
  benefit of making denial-of-service attacks considerably harder to



Cerf, et al.                 Informational                     [Page 24]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  mount more generally, as compared with conventional Internet routers.
  However, the obvious cost for this capability is potentially larger
  computation and credential storage overhead required at DTN nodes.

  For more detailed information on DTN security provisions, refer to
  [DTNSEC] and [DTNSOV].

4.  State Management Considerations

  An important aspect of any networking architecture is its management
  of state.  This section describes the state managed at the bundle
  layer and discusses how it is established and removed.

4.1.  Application Registration State

  In long/variable delay environments, an asynchronous application
  interface seems most appropriate.  Such interfaces typically include
  methods for applications to register callback actions when certain
  triggering events occur (e.g., when ADUs arrive).  These
  registrations create state information called application
  registration state.

  Application registration state is typically created by explicit
  request of the application, and is removed by a separate explicit
  request, but may also be removed by an application-specified timer
  (it is thus "firm" state).  In most cases, there must be a provision
  for retaining this state across application and operating system
  termination/restart conditions because a client/server bundle round-
  trip time may exceed the requesting application's execution time (or
  hosting system's uptime).  In cases where applications are not
  automatically restarted but application registration state remains
  persistent, a method must be provided to indicate to the system what
  action to perform when the triggering event occurs (e.g., restarting
  some application, ignoring the event, etc.).

  To initiate a registration and thereby establish application
  registration state, an application specifies an Endpoint ID for which
  it wishes to receive ADUs, along with an optional time value
  indicating how long the registration should remain active.  This
  operation is somewhat analogous to the bind() operation in the common
  sockets API.

  For registrations to groups (i.e., joins), a time interval may also
  be specified.  The time interval refers to the range of origination
  times of ADUs sent to the specified EID.  See Section 3.4 above for
  more details.





Cerf, et al.                 Informational                     [Page 25]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


4.2.  Custody Transfer State

  Custody transfer state includes information required to keep account
  of bundles for which a node has taken custody, as well as the
  protocol state related to transferring custody for one or more of
  them.  The accounting-related state is created when a bundle is
  received.  Custody transfer retransmission state is created when a
  transfer of custody is initiated by forwarding a bundle with the
  custody transfer requested delivery option specified.  Retransmission
  state and accounting state may be released upon receipt of one or
  more Custody Transfer Succeeded signals, indicating custody has been
  moved.  In addition, the bundle's expiration time (possibly mitigated
  by local policy) provides an upper bound on the time when this state
  is purged from the system in the event that it is not purged
  explicitly due to receipt of a signal.

4.3.  Bundle Routing and Forwarding State

  As with the Internet architecture, we distinguish between routing and
  forwarding.  Routing refers to the execution of a (possibly
  distributed) algorithm for computing routing paths according to some
  objective function (see [JFP04], for example).  Forwarding refers to
  the act of moving a bundle from one DTN node to another.  Routing
  makes use of routing state (the RIB, or routing information base),
  while forwarding makes use of state derived from routing, and is
  maintained as forwarding state (the FIB, or forwarding information
  base).  The structure of the FIB and the rules for maintaining it are
  implementation choices.  In some DTNs, exchange of information used
  to update state in the RIB may take place on network paths distinct
  from those where exchange of application data takes place.

  The maintenance of state in the RIB is dependent on the type of
  routing algorithm being used.  A routing algorithm may consider
  requested class of service and the location of potential custodians
  (for custody transfer, see section 3.10), and this information will
  tend to increase the size of the RIB.  The separation between FIB and
  RIB is not required by this document, as these are implementation
  details to be decided by system implementers.  The choice of routing
  algorithms is still under study.

  Bundles may occupy queues in nodes for a considerable amount of time.
  For unicast or anycast delivery, the amount of time is likely to be
  the interval between when a bundle arrives at a node and when it can
  be forwarded to its next hop.  For multicast delivery of bundles,
  this could be significantly longer, up to a bundle's expiration time.
  This situation occurs when multicast delivery is utilized in such a
  way that nodes joining a group can obtain information previously sent
  to the group.  In such cases, some nodes may act as "archivers" that



Cerf, et al.                 Informational                     [Page 26]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  provide copies of bundles to new participants that have already been
  delivered to other participants.

4.4.  Security-Related State

  The DTN security approach described in [DTNSEC], when used, requires
  maintenance of state in all DTN nodes that use it.  All such nodes
  are required to store their own private information (including their
  own policy and authentication material) and a block of information
  used to verify credentials.  Furthermore, in most cases, DTN nodes
  will cache some public information (and possibly the credentials) of
  their next-hop (bundle) neighbors.  All cached information has
  expiration times, and nodes are responsible for acquiring and
  distributing updates of public information and credentials prior to
  the expiration of the old set (in order to avoid a disruption in
  network service).

  In addition to basic end-to-end and hop-by-hop authentication, access
  control may be used in a DTN by one or more mechanisms such as
  capabilities or access control lists (ACLs).  ACLs would represent
  another block of state present in any node that wishes to enforce
  security policy.  ACLs are typically initialized at node
  configuration time and may be updated dynamically by DTN bundles or
  by some out of band technique.  Capabilities or credentials may be
  revoked, requiring the maintenance of a revocation list ("black
  list", another form of state) to check for invalid authentication
  material that has already been distributed.

  Some DTNs may implement security boundaries enforced by selected
  nodes in the network, where end-to-end credentials may be checked in
  addition to checking the hop-by-hop credentials.  (Doing so may
  require routing to be adjusted to ensure all bundles comprising each
  ADU pass through these points.)  Public information used to verify
  end-to-end authentication will typically be cached at these points.

4.5.  Policy and Configuration State

  DTN nodes will contain some amount of configuration and policy
  information.  Such information may alter the behavior of bundle
  forwarding.  Examples of policy state include the types of
  cryptographic algorithms and access control procedures to use if DTN
  security is employed, whether nodes may become custodians, what types
  of convergence layer (see Section 6) and routing protocols are in
  use, how bundles of differing priorities should be scheduled, where
  and for how long bundles and other data is stored, what status
  reports may be generated or at what rate, etc.





Cerf, et al.                 Informational                     [Page 27]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


5.  Application Structuring Issues

  DTN bundle delivery is intended to operate in a delay-tolerant
  fashion over a broad range of network types.  This does not mean
  there *must* be large delays in the network; it means there *may* be
  very significant delays (including extended periods of disconnection
  between sender and intended recipient(s)).  The DTN protocols are
  delay tolerant, so applications using them must also be delay
  tolerant in order to operate effectively in environments subject to
  significant delay or disruption.

  The communication primitives provided by the DTN architecture are
  based on asynchronous, message-oriented communication which differs
  from conversational request/response communication.  In general,
  applications should attempt to include enough information in an ADU
  so that it may be treated as an independent unit of work by the
  network and receiver(s).  The goal is to minimize synchronous
  interchanges between applications that are separated by a network
  characterized by long and possibly highly variable delays.  A single
  file transfer request message, for example, might include
  authentication information, file location information, and requested
  file operation (thus "bundling" this information together).
  Comparing this style of operation to a classic FTP transfer, one sees
  that the bundled model can complete in one round trip, whereas an FTP
  file "put" operation can take as many as eight round trips to get to
  a point where file data can flow [DFS02].

  Delay-tolerant applications must consider additional factors beyond
  the conversational implications of long delay paths.  For example, an
  application may terminate (voluntarily or not) between the time it
  sends a message and the time it expects a response.  If this
  possibility has been anticipated, the application can be "re-
  instantiated" with state information saved in persistent storage.
  This is an implementation issue, but also an application design
  consideration.

  Some consideration of delay-tolerant application design can result in
  applications that work reasonably well in low-delay environments, and
  that do not suffer extraordinarily in high or highly-variable delay
  environments.

6.  Convergence Layer Considerations for Use of Underlying Protocols

  Implementation experience with the DTN architecture has revealed an
  important architectural construct and interface for DTN nodes
  [DBFJHP04].  Not all underlying protocols in different protocol
  families provide the same exact functionality, so some additional
  adaptation or augmentation on a per-protocol or per-protocol-family



Cerf, et al.                 Informational                     [Page 28]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  basis may be required.  This adaptation is accomplished by a set of
  convergence layers placed between the bundle layer and underlying
  protocols.  The convergence layers manage the protocol-specific
  details of interfacing with particular underlying protocols and
  present a consistent interface to the bundle layer.

  The complexity of one convergence layer may vary substantially from
  another, depending on the type of underlying protocol it adapts.  For
  example, a TCP/IP convergence layer for use in the Internet might
  only have to add message boundaries to TCP streams, whereas a
  convergence layer for some network where no reliable transport
  protocol exists might be considerably more complex (e.g., it might
  have to implement reliability, fragmentation, flow-control, etc.) if
  reliable delivery is to be offered to the bundle layer.

  As convergence layers implement protocols above and beyond the basic
  bundle protocol specified in [BSPEC], they will be defined in their
  own documents (in a fashion similar to the way encapsulations for IP
  datagrams are specified on a per-underlying-protocol basis, such as
  in RFC 894 [RFC894]).

7.  Summary

  The DTN architecture addresses many of the problems of heterogeneous
  networks that must operate in environments subject to long delays and
  discontinuous end-to-end connectivity.  It is based on asynchronous
  messaging and uses postal mail as a model of service classes and
  delivery semantics.  It accommodates many different forms of
  connectivity, including scheduled, predicted, and opportunistically
  connected delivery paths.  It introduces a novel approach to end-to-
  end reliability across frequently partitioned and unreliable
  networks.  It also proposes a model for securing the network
  infrastructure against unauthorized access.

  It is our belief that this architecture is applicable to many
  different types of challenged environments.

8.  Security Considerations

  Security is an integral concern for the design of the Delay Tolerant
  Network Architecture, but its use is optional.  Sections 3.6.1, 3.14,
  and 4.4 of this document present some factors to consider for
  securing the DTN architecture, but separate documents [DTNSOV] and
  [DTNSEC] define the security architecture in much more detail.







Cerf, et al.                 Informational                     [Page 29]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


9.  IANA Considerations

  This document specifies the architecture for Delay Tolerant
  Networking, which uses Internet-standard URIs for its Endpoint
  Identifiers.  URIs intended for use with DTN should be compliant with
  the guidelines given in [RFC3986].

10.  Normative References

  [RFC3986]   Berners-Lee, T., Fielding, R., and L. Masinter, "Uniform
              Resource Identifier (URI): Generic Syntax", STD 66, RFC
              3986, January 2005.

11.  Informative References

  [IPN01]     InterPlaNetary Internet Project, Internet Society IPN
              Special Interest Group, http://www.ipnsig.org.

  [SB03]      S. Burleigh, et al., "Delay-Tolerant Networking - An
              Approach to Interplanetary Internet", IEEE Communications
              Magazine, July 2003.

  [FW03]      F. Warthman, "Delay-Tolerant Networks (DTNs): A Tutorial
              v1.1", Wartham Associates, 2003.  Available from
              http://www.dtnrg.org.

  [KF03]      K. Fall, "A Delay-Tolerant Network Architecture for
              Challenged Internets", Proceedings SIGCOMM, Aug 2003.

  [JFP04]     S. Jain, K. Fall, R. Patra, "Routing in a Delay Tolerant
              Network", Proceedings SIGCOMM, Aug/Sep 2004.

  [DFS02]     R. Durst, P. Feighery, K. Scott, "Why not use the
              Standard Internet Suite for the Interplanetary
              Internet?", MITRE White Paper, 2002.  Available from
              http://www.ipnsig.org/reports/TCP_IP.pdf.

  [CK74]      V. Cerf, R. Kahn, "A  Protocol for Packet Network
              Intercommunication", IEEE Trans. on Comm., COM-22(5), May
              1974.

  [IGE00]     C. Intanagonwiwat, R. Govindan, D. Estrin, "Directed
              Diffusion: A Scalable and Robust Communication Paradigm
              for Sensor Networks", Proceedings MobiCOM, Aug 2000.







Cerf, et al.                 Informational                     [Page 30]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  [WSBL99]    W. Adjie-Winoto, E. Schwartz, H. Balakrishnan, J. Lilley,
              "The Design and Implementation of an Intentional Naming
              System", Proc. 17th ACM SOSP, Kiawah Island, SC, Dec.
              1999.

  [CT90]      D. Clark, D. Tennenhouse, "Architectural Considerations
              for a New Generation of Protocols", Proceedings SIGCOMM,
              1990.

  [ISCHEMES]  IANA, Uniform Resource Identifer (URI) Schemes,
              http://www.iana.org/assignments/uri-schemes.html.

  [JDPF05]    S. Jain, M. Demmer, R. Patra, K. Fall, "Using Redundancy
              to Cope with Failures in a Delay Tolerant Network",
              Proceedings SIGCOMM, 2005.

  [WJMF05]    Y. Wang, S. Jain, M. Martonosi, K. Fall, "Erasure Coding
              Based Routing in Opportunistic Networks", Proceedings
              SIGCOMM Workshop on Delay Tolerant Networks, 2005.

  [ZAZ05]     W. Zhao, M. Ammar, E. Zegura, "Multicast in Delay
              Tolerant Networks", Proceedings SIGCOMM Workshop on Delay
              Tolerant Networks, 2005.

  [LFC05]     J. Leguay, T. Friedman, V. Conan, "DTN Routing in a
              Mobility Pattern Space", Proceedings SIGCOMM Workshop on
              Delay Tolerant Networks, 2005.

  [AF03]      J. Alonso, K. Fall, "A Linear Programming Formulation of
              Flows over Time with Piecewise Constant Capacity and
              Transit Times", Intel Research Technical Report IRB-TR-
              03-007, June 2003.

  [FHM03]     K. Fall, W. Hong, S. Madden, "Custody Transfer for
              Reliable Delivery in Delay Tolerant Networks", Intel
              Research Technical Report IRB-TR-03-030, July 2003.

  [BSPEC]     K. Scott, S. Burleigh, "Bundle Protocol Specification",
              Work in Progress, December 2006.

  [DTNSEC]    S. Symington, S. Farrell, H. Weiss, "Bundle Security
              Protocol Specification", Work in Progress, October 2006.

  [DTNSOV]    S. Farrell, S. Symington, H. Weiss, "Delay-Tolerant
              Networking Security Overview", Work in Progress, October
              2006.





Cerf, et al.                 Informational                     [Page 31]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  [DBFJHP04]  M. Demmer, E. Brewer, K. Fall, S. Jain, M. Ho, R. Patra,
              "Implementing Delay Tolerant Networking", Intel Research
              Technical Report IRB-TR-04-020, Dec. 2004.

  [RFC792]    Postel, J., "Internet Control Message Protocol", STD 5,
              RFC 792, September 1981.

  [RFC894]    Hornig, C., "A Standard for the Transmission of IP
              Datagrams over Ethernet Networks", STD 41, RFC 894, April
              1 1984.

  [RFC2960]   Stewart, R., Xie, Q., Morneault, K., Sharp, C.,
              Schwarzbauer, H., Taylor, T., Rytina, I., Kalla, M.,
              Zhang, L., and V. Paxson, "Stream Control Transmission
              Protocol", RFC 2960, October 2000.

  [RFC4088]   Black, D., McCloghrie, K., and J. Schoenwaelder, "Uniform
              Resource Identifier (URI) Scheme for the Simple Network
              Management Protocol (SNMP)", RFC 4088, June 2005.

  [S05]       K. Scott, "Disruption Tolerant Networking Proxies for
              On-the-Move Tactical Networks", Proc. MILCOM 2005
              (unclassified track), Oct. 2005.

  [T02]       W. Thies, et al., "Searching the World Wide Web in Low-
              Connectivity Communities", Proc. WWW Conference (Global
              Community track), May 2002.

12.  Acknowledgments

  John Wroclawski, David Mills, Greg Miller, James P. G. Sterbenz, Joe
  Touch, Steven Low, Lloyd Wood, Robert Braden, Deborah Estrin, Stephen
  Farrell, Melissa Ho, Ting Liu, Mike Demmer, Jakob Ericsson, Susan
  Symington, Andrei Gurtov, Avri Doria, Tom Henderson, Mark Allman,
  Michael Welzl, and Craig Partridge all contributed useful thoughts
  and criticisms to versions of this document.  We are grateful for
  their time and participation.

  This work was performed in part under DOD Contract DAA-B07-00-CC201,
  DARPA AO H912; JPL Task Plan No. 80-5045, DARPA AO H870; and NASA
  Contract NAS7-1407.










Cerf, et al.                 Informational                     [Page 32]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


Authors' Addresses

  Dr. Vinton G. Cerf
  Google Corporation
  Suite 384
  13800 Coppermine Rd.
  Herndon, VA 20171
  Phone: +1 (703) 234-1823
  Fax:   +1 (703) 848-0727
  EMail: [email protected]

  Scott C. Burleigh
  Jet Propulsion Laboratory
  4800 Oak Grove Drive
  M/S: 179-206
  Pasadena, CA 91109-8099
  Phone: +1 (818) 393-3353
  Fax:   +1 (818) 354-1075
  EMail: [email protected]

  Robert C. Durst
  The MITRE Corporation
  7515 Colshire Blvd., M/S H440
  McLean, VA 22102
  Phone: +1 (703) 983-7535
  Fax:   +1 (703) 983-7142
  EMail: [email protected]

  Dr. Kevin Fall
  Intel Research, Berkeley
  2150 Shattuck Ave., #1300
  Berkeley, CA 94704
  Phone: +1 (510) 495-3014
  Fax:   +1 (510) 495-3049
  EMail: [email protected]

  Adrian J. Hooke
  Jet Propulsion Laboratory
  4800 Oak Grove Drive
  M/S: 303-400
  Pasadena, CA 91109-8099
  Phone: +1 (818) 354-3063
  Fax:   +1 (818) 393-3575
  EMail: [email protected]







Cerf, et al.                 Informational                     [Page 33]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


  Dr. Keith L. Scott
  The MITRE Corporation
  7515 Colshire Blvd., M/S H440
  McLean, VA 22102
  Phone: +1 (703) 983-6547
  Fax:   +1 (703) 983-7142
  EMail: [email protected]

  Leigh Torgerson
  Jet Propulsion Laboratory
  4800 Oak Grove Drive
  M/S: 238-412
  Pasadena, CA 91109-8099
  Phone: +1 (818) 393-0695
  Fax:   +1 (818) 354-6825
  EMail: [email protected]

  Howard S. Weiss
  SPARTA, Inc.
  7075 Samuel Morse Drive
  Columbia, MD 21046
  Phone: +1 (410) 872-1515 x201
  Fax:   +1 (410) 872-8079
  EMail: [email protected]

  Please refer comments to [email protected].  The Delay
  Tolerant Networking Research Group (DTNRG) web site is located at
  http://www.dtnrg.org.























Cerf, et al.                 Informational                     [Page 34]

RFC 4838         Delay-Tolerant Networking Architecture       April 2007


Full Copyright Statement

  Copyright (C) The IETF Trust (2007).

  This document is subject to the rights, licenses and restrictions
  contained in BCP 78, and except as set forth therein, the authors
  retain all their rights.

  This document and the information contained herein are provided on an
  "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
  OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND
  THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS
  OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF
  THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
  WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Intellectual Property

  The IETF takes no position regarding the validity or scope of any
  Intellectual Property Rights or other rights that might be claimed to
  pertain to the implementation or use of the technology described in
  this document or the extent to which any license under such rights
  might or might not be available; nor does it represent that it has
  made any independent effort to identify any such rights.  Information
  on the procedures with respect to rights in RFC documents can be
  found in BCP 78 and BCP 79.

  Copies of IPR disclosures made to the IETF Secretariat and any
  assurances of licenses to be made available, or the result of an
  attempt made to obtain a general license or permission for the use of
  such proprietary rights by implementers or users of this
  specification can be obtained from the IETF on-line IPR repository at
  http://www.ietf.org/ipr.

  The IETF invites any interested party to bring to its attention any
  copyrights, patents or patent applications, or other proprietary
  rights that may cover technology that may be required to implement
  this standard.  Please address the information to the IETF at
  [email protected].

Acknowledgement

  Funding for the RFC Editor function is currently provided by the
  Internet Society.







Cerf, et al.                 Informational                     [Page 35]


You are viewing proxied material from gopher.rbfh.de. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.