[Docs] [txt|pdf] [Tracker] [WG] [Email] [Diff1] [Diff2] [Nits]

Versions: 00 01

Network Working Group                                           Q. Xie
INTERNET-DRAFT                                                Motorola
                                                         R. R. Stewart
                                                              C. Sharp
                                                                 Cisco
                                                             I. Rytina
                                                              Ericsson
Expires in six months                                    February 2001


                   SCTP Unreliable Data Mode Extension
                    <draft-ietf-sigtran-usctp-01.txt>

Status of This Memo

This document is an Internet-Draft and is in full conformance with all
provisions of Section 10 of RFC 2026 [RFC2026]. Internet-Drafts are
working documents of the Internet Engineering Task Force (IETF), its
areas, and its working groups. Note that other groups may also
distribute working documents as Internet-Drafts.

The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt

The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.




Abstract

This document describes an extension to the Stream Control
Transmission Protocol (SCTP) [RFC2960] to provide unreliable data
transfer services. The benefits of this extension includes unified
congestion control over reliable and unreliable data streams, single
association for multi-content data services, link level
fault tolerance for unreliable data transfer, unreliable data stream
multiplexing, etc.

The data transfer service extension will also support partial payload
checksum in order to facilitate bit-error tolerance media
applications.


                    TABLE OF CONTENTS
1. Introduction................................................. 2
2. Conventions.................................................. 3
3. Unreliable Data and Partial Checksum Design.................. 3
  3.1 New INIT and INIT-ACK parameters.......................... 3
    3.1.1 Unreliable Streams Parameter Definition............... 3
    3.1.2 Partial Checksum Parameter Definition................. 5
  3.2 New chunk definitions..................................... 5
    3.2.1 Forward Cumulative TSN Chunk (FORWARD TSN)............ 5
    3.2.2 Partial Checksum Data Chunk (P-DATA).................. 6

Xie, et al                                                      [Page 1]


Internet Draft     SCTP Unreliable Data Mode Extension     February 2001

4. Unreliable Stream Operations................................. 8
  4.1 Initialization of Unreliable Streams...................... 8
  4.2 Send Unreliable Data...................................... 9
  4.3 Receive Unreliable Data...................................11
  4.4 Other Issues on Unreliable Data...........................11
    4.4.1 Unreliable Data Stream Multiplexing...................11
    4.4.2 Fault Tolerant Unreliable Data Transfer...............12
    4.4.3 Unreliable Data Out-of-order Detection................12
5. Usage of the P-DATA chunk....................................12
6. Acknowledgments..............................................13
7. Authors' Addresses...........................................13
8. References...................................................14


1. Introduction

Taking advantage of the extensibility of SCTP, this document adds
unreliable data transfer services to SCTP and an optional method to
send SCTP Data Chunks with limited checksum coverage. The design
presented here allows the co-existence of unreliable data streams
and reliable streams in a single SCTP association.

The following are some of the advantages for integrating unreliable
and partial checksum data services into SCTP:

  1) Unreliable extension to SCTP (U-SCTP) supports congestion
     control and congestion avoidance over unreliable data traffic;
     this is very desirable since it is much friendlier towards the
     network than UDP.

  2) Some applications services can greatly benefit from U-SCTP by
     using a single SCTP association to carry both reliable content
     (e.g., text, billing, accounting, set-up information, etc.) and
     unreliable content (e.g., Fiber channel, SCSI over IP, etc.).

  3) U-SCTP allows the use of a unified congestion control across
     both reliable and unreliable traffic between two endpoints. This
     has the potential for better utilization of network resources,
     achieving similar objectives of the Endpoint Congestion
     Management (ecm) Working Group.

  4) Taking advantage of SCTP data chunk bundling function, sending
     multiple unreliable data streams across a single SCTP association
     creates a very efficient and effective way of data multiplexing.

  5) U-SCTP gives even the unreliable data traffic "link-level" fault
     tolerance, taking advantage of SCTP multi-homing ability. This is
     not possible with UDP.

  6) U-SCTP can achieve either ordered or unordered unreliable data
     transfer, while UDP is incapable of controlling the order of data
     delivery.

  7) An application can control its retransmission policies if

Xie, et al                                                      [Page 2]


Internet Draft     SCTP Unreliable Data Mode Extension     February 2001

     retransmission is deemed needed.

  8) Some applications may find it desirable to limit the coverage
     of the Adler32 checksum over the actual data chunks.


2. Conventions

The keywords MUST, MUST NOT, REQUIRED, SHALL, SHALL NOT, SHOULD,
SHOULD NOT, RECOMMENDED, NOT RECOMMENDED, MAY, and OPTIONAL, when they
appear in this document, are to be interpreted as described in
RFC 2119 [RFC2119].


3. Unreliable Data and Partial Checksum Design

With the unreliable data extension, an SCTP data sender will be allowed to
designate a sub-set of its outbound streams to be unreliable
streams. The user data chunks sent to an unreliable stream will share
the same TSN space, the same congestion control/avoidance treatment,
and the same transmission priority as those sent to a reliable stream,
but they will not be retransmitted (or only be retransmitted for a
limited times) if they are found missing at the data receiver.

The partial checksum extension allows the sending application to
specify a portion of an outbound user message to be excluded from SCTP
checksum protection.

3.1 New INIT and INIT-ACK parameters

The following new optional parameter, are added to the INIT
and INIT ACK chunks. At the initialization of the association, the
sender of the INIT or INIT ACK chunk may include these parameters
to indicate its ability to support these features.

Parameter Name                       Status     Type Value
-------------------------------------------------------------
Unreliable Streams                  Optional    0xC000
Partial Checksum Support            Optional    0xC004


3.1.1 Unreliable Streams Parameter Definition

The Unreliable Streams parameter is added to the INIT and INIT ACK
chunks. At the initialization of the association, the sender of the
INIT or INIT ACK chunk shall include this optional parameter
to inform its peer that it is able to support unreliable streams
and to designate its unreliable outbound streams. If no streams
are marked as unreliable but the sender does support the
unreliable streams option the sender SHOULD include a parameter
with no u-stream ranges and a fixed Parameter Length of 4.


   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

Xie, et al                                                      [Page 3]


Internet Draft     SCTP Unreliable Data Mode Extension     February 2001

  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |    Parameter Type = 0xC000    |     Parameter Length          |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |      u-stream start #1 = US1  |      u-stream end #1 = UE1    |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  /                                                               /
  \                            . . . .                            \
  /                                                               /
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |      u-stream start #k = USk  |       u-stream end #k = UEk   |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  Type: 16 bit u_int

     0xC000, indicating Unreliable Streams parameter

  Length: 16 bit u_int

     Indicate the size of the parameter in octets, including the
     Type, Length, u-stream start, and u-stream end fields.

  u-stream start: 16 bit u_int, and
  u-stream end: 16 bit u_int

     Each pair of u-stream start and u-stream end fields defines one or more
     unreliable outbound streams, starting from stream number US and
     ending with stream number UE. The union of all the pairs
     together defines the complete sub-set of all unreliable
     outbound streams.

     The following are some examples of unreliable stream designation
     (assuming OS = 10):

     Example 1: (assuming OS = 10)

       +------------+-----------+
       |type=0xC000 | length=8  |         Streams    Mode
       +------------+-----------+    ==>  ---------------------
       | u-start= 3 | u-end= 5  |         0 - 2      reliable
       +------------+-----------+         3 - 5      unreliable
                                          6 - 9      reliable

     Example 2: (assuming OS = 10)

       +------------+-----------+
       |type=0xC000 | length=12 |         Streams    Mode
       +------------+-----------+    ==>  ---------------------
       | u-start= 3 |  u-end= 5 |         0 - 2      reliable
       +------------+-----------+         3 - 9     unreliable
       | u-start= 6 |  u-end= 9 |
       +------------+-----------+

     Example 3: (assuming OS = 10)


Xie, et al                                                      [Page 4]


Internet Draft     SCTP Unreliable Data Mode Extension     February 2001

       +------------+-----------+
       |type=0xC000 | length=12 |         Streams    Mode
       +------------+-----------+    ==>  ---------------------
       | u-start= 9 |  u-end= 9 |         0          unreliable
       +------------+-----------+         1 - 8      reliable
       | u-start= 0 |  u-end= 0 |         9          unreliable
       +------------+-----------+

     Example 4: (assuming OS = 10)

       +------------+-----------+
       |type=0xC000 | length=8  |         Streams    Mode
       +------------+-----------+    ==>  ---------------------
       | u-start= 0 |  u-end= 9 |         0 - 9      unreliable
       +------------+-----------+

     Example 5: (assuming OS = 10)
       +------------+-----------+
       |type=0xC000 | length=4  |         Streams    Mode
       +------------+-----------+    ==>  ---------------------
                                          0 - 9      reliable


3.1.2  Partial Checksum Parameter Definition

The Partial Checksum Parameter is added to the INIT and INIT ACK
chunks. At the initialization of the association, the sender of the
INIT or INIT ACK chunk shall include this optional parameter
to inform its peer that it is able to support the partial checksum
P-DATA (Chunk Type = 0xC3 or 195), see section 3.2.2.


   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
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |    Parameter Type = 0xc004    |     Parameter Length=4        |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+


3.2 New chunk definitions

The following new chunk types are added to support the two new
features in SCTP.

  Chunk Type    Chunk Name
  ------------------------------------------------------
  11000000      Forward Cumulative TSN (FORWARD TSN)
  11000011      Partial Checksum Data Chunk (P-DATA)

The FORWARD TSN supports the unreliable stream. The Partial Checksum
Data Chunk supports data chunks that are not completely covered by the
Adler32 checksum in the SCTP packet header.


3.2.1 Forward Cumulative TSN Chunk Definition (FORWARD TSN)

Xie, et al                                                      [Page 5]


Internet Draft     SCTP Unreliable Data Mode Extension     February 2001


Forward Cumulative TSN chunk shall be used by the data sender to
inform the data receiver to adjust its cumulative received TSN point
forward because some missing TSNs are associated with unreliable data
chunks and will no longer be retransmitted by the sender.

   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
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |1 1 0 0 0 0 0 0|  Chunk Flags  |0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0|
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  |                      New Cumulative TSN                       |
  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

Chunk Flags:

  Set to all zeros on transmit and ignored on receipt.

New Cumulative TSN: 32 bit u_int

  This indicates the new cumulative TSN to the data receiver.
  Upon the reception of this value, the data receiver shall consider
  any missing TSNs earlier than this value received and stop reporting
  them as gaps in the subsequent SACKs.


3.2.2 Partial Checksum Data Chunk (P-DATA)

The following format MUST be used for the P-DATA chunk:

    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
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |1 1 0 0 0 0 1 1| Reserved|U|B|E|    Length                     |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                              TSN                              |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |      Stream Identifier S      |   Stream Sequence Number n    |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                 Payload Protocol Identifier                   |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                      Checksum Coverage                        |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   \                                                               \
   /                 User Data (seq n of Stream S)                 /
   \                                                               \
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

Reserved: 5 bits

  Should be set to all '0's and ignored by the receiver.

U bit: 1 bit


Xie, et al                                                      [Page 6]


Internet Draft     SCTP Unreliable Data Mode Extension     February 2001

  Same as in a DATA chunk [RFC2960], the (U)nordered bit, if set to '1',
  indicates that this is an unordered P-DATA chunk, and there is no
  Stream Sequence Number assigned to this P-DATA chunk. Therefore, the
  receiver MUST ignore the Stream Sequence Number field.

  Unordered P-DATA chunks MUST be dispatched to the upper layer by the
  receiver without any attempt to re-order them.

B bit: 1 bit

  MUST be set to 1 by the sender.

E bit: 1 bit

  MUST be set to 1 by the sender.

Length: 16 bits (unsigned integer)

  This field indicates the length of the P-DATA chunk in bytes from the
  beginning of the type field to the end of the user data field
  excluding any padding. For example, a P-DATA chunk with 5 bytes of
  user data field will have Length set to 25 (indicating 25 bytes).

TSN: 32 bits (unsigned integer)

  This value represents the TSN for this P-DATA chunk. The valid range
  of TSN is from 0 to 4294967295 (2**32 - 1).  TSN wraps back to 0
  after reaching 4294967295.

Stream Identifier S: 16 bits (unsigned integer)

  Identifies the stream to which the following the user data belongs.

Stream Sequence Number n: 16 bits (unsigned integer)

  This value represents the stream sequence number of the following
  user data within the stream S. Valid range is 0 to 65535.

Payload Protocol Identifier: 32 bits (unsigned integer)

  Same as in a DATA chunk [RFC2960], this value represents an
  application (or upper layer) specified protocol identifier.  This
  value is passed to SCTP by its upper layer and sent to its peer.
  This identifier is not used by SCTP but can be used by certain
  network entities as well as the peer application to identify the
  type of information being carried in this P-DATA chunk.

  The value 0 indicates no application identifier is specified by
  the upper layer for this payload data.

Checksum Coverage: 32 bits (unsigned integer)

  This field contains an integer that indicates how much User Data (in
  octets) of this P-DATA chunk is covered by the Adler32 checksum in

Xie, et al                                                      [Page 7]


Internet Draft     SCTP Unreliable Data Mode Extension     February 2001

  the SCTP packet common header.

    Note, all chunk headers, as well as the common header of the SCTP
    packet are always covered by the packet checksum, regardless of
    the value of the Checksum Coverage.

  The coverage area is defined as starting from the first transmitted
  byte in the User Data part of the Chunk for exactly
  'Checksum Coverage' bytes in length.

  For example, if a value of "5" appears in the Checksum Coverage
  field, only the first 5 bytes of the User Data are included in the
  calculation of the packet checksum. A value of "0" in the Checksum
  Coverage field will indicate that all the User Data in this P-DATA
  chunk is excluded from the packet checksum.

  The sender of a P-DATA chunk SHOULD NOT set a value of Checksum
  Coverage larger than the length of the User Data in the chunk. In
  other words, the Checksum Coverage SHOULD be set smaller than or
  equal to the chunk Length - 20.

  On the receiver side, if the value of Checksum Coverage in a
  received P-DATA chunk is found larger than the length of the User
  Data in the chunk, the receiver MUST accept the chunk and MUST
  consider the entire User Data in the chunk covered by the Adler-32
  checksum.

User Data: variable length

  This is the payload user data. The implementation MUST pad the end
  of the data to a 4 byte boundary with all-zero bytes. Any padding
  MUST NOT be included in the Length field. A sender MUST never add
  more than 3 bytes of padding.


4. Unreliable Stream Operations

In this section, we first defines the procedures for opening
unreliable streams in an SCTP association. Then, we will discuss
procedures for sending and receiving unreliable SCTP data chunks.


4.1 Initialization of Unreliable Streams

If the SCTP data sender plans to send unreliable data, at the
initialization of the association it MUST include the Unreliable
Streams parameter in its INIT or INIT ACK chunk to indicate to its
peer which of its outbound streams are going to be used as unreliable
streams.

Upon the reception of the Unreliable Streams parameter, the data
receiver SHALL determine and record the mode (reliable or unreliable)
of each inbound stream, as it allocates resource for its inbound
streams.

Xie, et al                                                      [Page 8]


Internet Draft     SCTP Unreliable Data Mode Extension     February 2001


Note, if the data receiver does not support unreliable inbound
streams, it SHOULD treat the Unreliable Streams parameter as an
invalid or unrecognized parameter and respond to the data sender with
an operational error, following the rules defined in Section 5.1
of [RFC2960].

Upon reception of the operational error indicating that its peer does
not support unreliable streams, the data sender may choose to either:

  1) end the initiation process, in consideration of the peer's
     inability of meeting the requested features for the new
     association, or
  2) continue the initiation process, but with the understanding that
     ALL its outbound streams will be reliable.

In either case, the data sender SHOULD inform its upper layer its
peer's inability of supporting unreliable data transfer.

Initiation of streams as reliable and/or unreliable may be under the
control of the SCTP user. Hence, the ULP primitive "ASSOCIATE" (see
Section 10.1 of [RFC2960]) should be expanded to contain the optional
U-stream-start and U-stream-end values.


4.2 Send Unreliable Data

During the lifetime of the association, any user data sent to an
unreliable stream will be treated as unreliable user data and will
automatically be transmitted in unreliable mode.

Note, fragmentation MUST NOT be performed on an unreliable user
message. In other words, an unreliable user message MUST be sent in a
single DATA or P-DATA chunk regardless of its size.

The SCTP data sender shall handle user data sent to an unreliable
stream the same way as it handles user data sent to a reliable stream
(i.e., the same timer rules, congestion control rules, failure
detection rules, RTO control rules, etc.), with the following
exceptions:

 A1) The sender maintains a "Peer.Cumulative.TSN" for each peer so as
     to track the latest cumulative TSN point of the peer (Note, this
     variable may already exist in a classic SCTP implementation for
     outqueue management and for detecting out-of-order SACKs).

 A2) Before retransmitting a DATA chunk (due to either a T3-rtx timer
     expiration as defined in 6.3.3 of [RFC2960] or a 4th missing
     indication as defined in 7.2.4 of [RFC2960]), the SCTP data
     sender MUST check whether the DATA chunk is being transmitted on
     an unreliable stream. If so, it will perform the following:

     B1) Check the value of the unreliable retransmission counter
         "Unrel.Trans.Count" value for the DATA chunk. This value may

Xie, et al                                                      [Page 9]


Internet Draft     SCTP Unreliable Data Mode Extension     February 2001

         be set by the SCTP user to 0 (no retransmission) for complete
         unreliability, or N (where N >0) for limited reliability at
         the time when the user message is passed to SCTP.

     B2) If the "Unrel.Trans.Count" of the chunk is currently greater
         than 0, the sender MUST retransmit the data chunk and then
         decrease the "Unrel.Trans.Count" by 1. The same rules for
         retransmission as defined in [RFC2960] SHALL be used for RTO
         calculation, destination selection, error reporting, etc.

     B3) If the "Unrel.Trans.Count" is currently 0, the sender MUST NOT
         retransmit the data chunk. Instead, the sender MUST mark the
         data chunk as being finally acked.

 A3) whenever the data sender receives a SACK from the data receiver,
     it SHALL first process the SACK using the normal procedures as
     defined in [RFC2960] for classic SCTP. This includes, among other
     things: a) check if the SACK is received out-of-order; if not, b)
     adopt the Cumulative TSN ACK carried in the SACK as the new
     "Peer.Cumulative.TSN"; and c) process any gap reports if present
     (see [RFC2960]).

     The data sender MUST then perform the following additional
     steps:

     C1) Try to further advance the "Peer.Cumulative.TSN" locally,
         that is, to move "Peer.Cumulative.TSN" up as long as the
         chunk next in the out-queue is marked as acknowledged. For
         example (assuming a SACK arrived with the Cumulative TSN
         ACK = 102),

          out-queue at the end of  ==>   out-queue after cmTSN
          normal SACK processing           local advancement

                  ...                           ...
         cmTSN -> 102 acked                     102 acked
                  103 acked                     103 acked
                  104 acked            cmTSN -> 104 acked
                  105                           105
                  106 acked                     106 acked
                  ...                           ...

         Here, the data sender successfully advanced the
         "Peer.Cumulative.TSN" from 102 to 104 locally.

     C2) Whenever a local advancement of the "Peer.Cumulative.TSN" is
         made, the data sender MUST send the data receiver a
         FORWARD TSN chunk containing the new value of the
         "Peer.Cumulative.TSN" (which is 104 in the above example).

         Note, an endpoint MUST NOT use the forward TSN for any other
         purposes other than the above circumstance.

     Note, if a TSN is indicated as missing by a SACK carrying gap

Xie, et al                                                     [Page 10]


Internet Draft     SCTP Unreliable Data Mode Extension     February 2001

     reports AND the TSN is earlier than the current
     "Peer.Cumulative.TSN", the data sender MUST NOT take any action
     on this TSN, i.e., it MUST ignore this missing report to this
     TSN. When this happens, it is normally an indication that a
     previous FORWARD TSN from the data sender may have been lost in
     the network.

     Note, the detection criterion for out-of-order SACKs MUST remains
     the same as stated in RFC2960, that is, a SACK is only considered
     out-of-order if the Cumulative TSN ACK carried in the SACK is
     earlier than that of the previous received SACK (i.e., the
     comparison MUST NOT be made against the locally advanced
     "Peer.Cumulative.TSN").

The ULP primitive "DATA" (defined in Section 10.1 of [RFC2960]) should
be expanded to contain an optional unreliable retransmission parameter
to assign a "Unrel.Trans.Count" value to each user message to be sent
to an unreliable stream.


4.3 Receive Unreliable Data

Regardless whether a DATA chunk arrives from a reliable stream or an
unreliable stream, the receiver MUST perform the same TSN handling
(e.g, duplicate detection, gap detection, SACK generation, cumulative
TSN advancement, etc.) as defined in [RFC2960].

However, whenever a FORWARD TSN chunk arrives the data receiver MUST
update its cumulative TSN to the value carried in the FORWARD TSN
chunk, and MUST stop reporting any missing TSNs before the new
cumulative TSN.

Whenever an unreliable DATA chunk arrives with the 'U' bit set to '0'
(indicating ordered delivery) and is out of order, the receiver must
hold the chunk for reordering. However since it is possible that the
DATA chunk(s) being waited upon is one that will not be retransmitted
by the sender, when a FORWARD TSN chunk arrives, the receiver MUST
examine all of its unreliable stream reordering queues, and
immediately make available for delivery any messages that carry a TSN
earlier than the new cumulative TSN updated by the FORWARD TSN.


4.4. Other Issues on Unreliable Data

4.4.1 Unreliable Data Stream Multiplexing

Sometimes, it is desirable to aggregate different real time media
streams (e.g., RTP streams) and send them over a single communication
connection, and normally unreliable transport is preferred for these
types of media streams.

With U-SCTP this is easily achieved by assigning each different media
stream to a different unreliable SCTP stream and enabling the SCTP
data bundling to perform the multiplexing.

Xie, et al                                                     [Page 11]


Internet Draft     SCTP Unreliable Data Mode Extension     February 2001


The implementation of the data sender MAY use a bundling timer with a
time interval adjusted to the timing characteristics of the specific
media type in order to achieve the optimal multiplexing efficiency.


4.4.2 Fault Tolerant Unreliable Data Transfer

When the data receiver is multi-homed, unreliable data transfer using
U-SCTP will obtain the same fault tolerance benefit as that of the
reliable data services across an SCTP association.

This is because the data sender still follows the same failure
detection rules and still counts the omitted retransmission against
the association and the destination transport address to which the
unreliable DATA chunk was originally sent. Thus, when failure occurs,
the data sender will detect the failure and shift the unreliable data
services to an alternate destination address, following the same
procedures as defined in Section 8 of [RFC2960] for reliable data transfer.


4.4.3 Unreliable Data Out-of-order Detection

Detecting out-of-order data in an unreliable stream is useful for some
applications (e.g. Fiber channel or SCSI over IP). With U-SCTP this
becomes possible - the upper layer simply needs to examine the the
stream sequence number of the arrived user messages to detect any
missing data or out-of-order data. This detection only works when
the DATA chunks are sent in order, i.e. their "U" bit MUST NOT be
set.


5 Usage of the P-DATA chunk.

For some applications it is beneficial NOT to discard a data
packet due to bit errors within the User Data portion. For this
type of applications this new optional chunk type is defined.

All rules defined in [RFC2960] for DATA Chunks MUST be followed for
the P-DATA chunk with the following exceptions:

 E1) When passing a partial checksum user message to SCTP data sender,
     the upper layer should indicate how many bytes of the first
     portion of the user message need checksum protection. The data
     sender will then put this value into the Checksum Coverage field
     of the outbound P-DATA chunk.

 E2) The Checksum Coverage field defines how much of this P-DATA
     chunk the Adler-32 checksum is to cover. During checksum
     computation the sender and receiver MUST use this field
     to determine how much of a P-DATA chunk to add to the
     Adler-32 checksum of the SCTP packet.

     For each P-DATA in a received SCTP packet, after summing the

Xie, et al                                                     [Page 12]


Internet Draft     SCTP Unreliable Data Mode Extension     February 2001

     chunk header bytes and the specified number of User Data bytes to
     the checksum, the receiver MUST skip to the next chunk if this
     P-DATA is not the last chunk in the SCTP packet and MUST NOT
     include the rest of the User Data in the P-DATA chunk in its
     checksum computation.

 E3) A sender MUST NOT send a P-DATA chunk unless the 'Partial Checksum
     Support' parameter was seen in the INIT or INIT-ACK from its peer.

It is important to note that P-DATA chunk uses the SAME TSN values as
the normal DATA Chunk type. The sender and receiver do NOT hold
separate TSN sequence spaces for DATA and P-DATA. Instead, the two
chunk types use a single TSN sequence space.

In effect the P-DATA chunk is treated in all considerations to be a
DATA chunk, with all of the normal DATA chunk rules for congestion
control effecting this chunk. The only difference in treatment of
P-DATA chunk comes during the calculation and verification of the
Adler-32 checksum.

P-DATA chunk MAY be used with either a reliable or un-reliable stream,
no restrictions are placed on its usage except those listed above.

Use of the P-DATA chunk may be under the control of the SCTP user.
Hence, the ULP primitive "DATA" (see section 10.1 of RFC2960) should
be expanded to contain an optional Checksum Coverage value.

When a reliable user message subscribing for partial checksum
protection is fragmented at the SCTP sender, the sender shall
calculate the Checksum Coverage value for each of the resulting P-DATA
chunks, based upon the user's original coverage requirement.


6. Acknowledgments

The authors would like to thank Scott Bradner, Jon Berger, and others
for their comments.


7. Authors' Addresses

Qiaobing Xie                            Tel: +1-847-632-3028
Motorola, Inc.                          EMail: qxie1@email.mot.com
1501 W. Shure Drive, #2309
Arlington Heights, IL 60004
USA

Randall R. Stewart                      Tel: +1-815-477-2127
Cisco Systems, Inc.                     EMail: rrs@cisco.com
8725 West Higgins Road
Suite 300
Chicago, Ill 60631

Chip Sharp                              Tel: +1-919-392-3121

Xie, et al                                                     [Page 13]


Internet Draft     SCTP Unreliable Data Mode Extension     February 2001

Cisco Systems Inc.                      EMail:chsharp@cisco.com
7025 Kit Creek Road
Research Triangle Park, NC  27709
USA

Ian Rytina                              Tel: +61-3-9301-6164
Ericsson Australia                      EMail:ian.rytina@ericsson.com
37/360 Elizabeth Street
Melbourne, Victoria 3000
Australia


8. References

[RFC2026] Bradner, S., "The Internet Standards Process -- Revision 3",
          RFC 2026, October 1996.

[RFC2119] Bradner, S. "Key words for use in RFCs to Indicate
          Requirement Levels", BCP 14, RFC 2119, March 1997.

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



      This Internet Draft expires in 6 months from February 2001.



























Xie, et al                                                     [Page 14]


Html markup produced by rfcmarkup 1.129c, available from https://tools.ietf.org/tools/rfcmarkup/