SIP compression: Compressing the Session Initiation Protocol (SIP)

Network Working Group                                       G. Camarillo
Request for Comments: 3486                                      Ericsson
Category: Standards Track                                  February 2003


           Compressing the Session Initiation Protocol (SIP)

Status of this Memo

   This document specifies an Internet standards track protocol for the
   Internet community, and requests discussion and suggestions for
   improvements.  Please refer to the current edition of the "Internet
   Official Protocol Standards" (STD 1) for the standardization state
   and status of this protocol.  Distribution of this memo is unlimited.

Copyright Notice

   Copyright (C) The Internet Society (2003).  All Rights Reserved.

Abstract

   This document describes a mechanism to signal that compression is
   desired for one or more Session Initiation Protocol (SIP) messages.
   It also states when it is appropriate to send compressed SIP messages
   to a SIP entity.

Table of Contents

   1.   Introduction ...............................................  2
   2.   Overview of operation ......................................  3
   3.   SigComp implementations for SIP ............................  3
   4.   Sending a Request to a Server ..............................  3
        4.1   Obtaining a SIP or SIPS URI with comp=sigcomp ........  4
   5.   Sending a Response to a Client .............................  5
   6.   Double Record-Routing ......................................  6
   7.   Error Situations ...........................................  6
   8.   Augmented BNF ..............................................  7
   9.   Example ....................................................  7
   10.  Security Considerations .................................... 10
   11.  IANA Considerations ........................................ 10
   12.  Acknowledgements............................................ 10
   13.  Normative References ....................................... 10
   14.  Informative References ..................................... 11
   15.  Author's Address............................................ 11
   16.  Full Copyright Statement.................................... 12

1.   Introduction

   A SIP [1] client sending a request to a SIP server typically performs
   a DNS lookup for the domain name of the server.  When NAPTR [4] or
   SRV [5] records are available for the server, the client can specify
   the type of service it wants.  The service in this context is the
   transport protocol to be used by SIP (e.g., UDP, TCP or SCTP).  A SIP
   server that supports, for instance, three different transport
   protocols, will have three different DNS entries.

   Since it is foreseen that the number of transport protocols supported
   by a particular application layer protocol is not going to grow
   dramatically, having a DNS entry per transport seems like a scalable
   enough solution.

   However, sometimes it is necessary to include new layers between the
   transport protocol and the application layer protocol.  Examples of
   these layers are transport layer security and compression.  If DNS
   was used to discover the availability of these layers for a
   particular server, the number of DNS entries needed for that server
   would grow dramatically.

   A server that, for example, supported TCP and SCTP as transports, TLS
   for transport security and SigComp for signaling compression, would
   need the 8 DNS entries listed below:

      1.   TCP, no security, no compression

      2.   TCP, no security, SigComp

      3.   TCP, TLS, no compression

      4.   TCP, TLS, SigComp

      5.   SCTP, no security, no compression

      6.   SCTP, no security, SigComp

      7.   SCTP, TLS, no compression

      8.   SCTP, TLS, SigComp

   It is clear that this way of using DNS is not scalable.  Therefore,
   an application layer mechanism to express support of signalling
   compression is needed.

      Note that for historical reasons both HTTP and SIP use a different
      port for TLS on top of TCP than for TCP alone, although at
      present, this solution is not considered scalable any longer.

   A SIP element that supports compression will need to be prepared to
   receive compressed and uncompressed messages on the same port.  It
   will perform demultiplexing based on the cookie in the topmost bits
   of every compressed message.

2.   Overview of operation

   There are two types of SIP messages; SIP requests and SIP responses.
   Clients send SIP requests to the host part of a URI and servers send
   responses to the host in the sent-by parameter of the Via header
   field.

   We define two parameters, one for SIP URIs and the other for the Via
   header field.  The format of both parameters is the same, as shown in
   the examples below:

   sip:alice@atlanta.com;comp=sigcomp
   Via: SIP/2.0/UDP server1.foo.com:5060;branch=z9hG4bK87a7;comp=sigcomp

   The presence of this parameter (comp=sigcomp) in a URI indicates that
   the request has to be compressed using SigComp, as defined in [2].
   The presence of comp=sigcomp in a Via header field indicates that the
   response has to be compressed using SigComp.

   Therefore, the presence of comp=sigcomp indicates that the SIP entity
   identified by the URI or by the Via header field supports SigComp and
   is willing to receive compressed messages.  Having comp=sigcomp mean
   "willingness" as well as "support" allows the receiver of a SIP
   message to influence the decision of whether or not to use SigComp at
   a given time.

3.   SigComp implementations for SIP

   Every SIP implementation that supports SigComp MUST implement the
   procedures described in this document.

4.   Sending a Request to a Server

   A request is sent to the host part of a URI.  This URI, referred to
   as the next-hop URI, is the Request-URI of the request or an entry in
   the Route header field.

   If the next-hop URI contains the parameter comp=sigcomp, the client
   SHOULD compress the request using SigComp as defined in [2].

   If the next-hop URI is a SIPS URI, the request SHOULD be compressed
   before it is passed to the TLS layer.

   A client MUST NOT send a compressed request to a server if it does
   not know whether or not the server supports SigComp.

   Regardless of whether the request is sent compressed or not, if a
   client would like to receive subsequent requests within the same
   dialog in the UAS->UAC direction compressed, this client SHOULD add
   the parameter comp=sigcomp to the URI in the Contact header field if
   it is a user agent client.  If the client is a proxy, it SHOULD add
   the parameter comp=sigcomp to its URI in the Record-Route header
   field.

   If a user agent client sends a compressed request, it SHOULD add the
   parameter comp=sigcomp to the URI in the Contact header field.  If a
   proxy that Record-Routes sends a compressed request, it SHOULD add
   comp=sigcomp to its URI in the Record-Route header field.

   If a client sends a compressed request, it SHOULD add the parameter
   comp=sigcomp to the topmost entry of the Via header field.

   If a client does not know whether or not the server supports SigComp,
   but in case the server supported it, it would like to receive
   compressed responses, this client SHOULD add the parameter
   comp=sigcomp to the topmost entry of the Via header field.  The
   request, however, as stated above, will not be compressed.

4.1   Obtaining a SIP or SIPS URI with comp=sigcomp

   For requests within a dialog, a next-hop URI with the comp=sigcomp
   parameter is obtained from a Record-Route header field when the
   dialog is established.  A client sending a request outside a dialog
   can also obtain SIP URIs with comp=sigcomp in a Contact header field
   in a 3xx or 485 response to the request.

   However, clients establishing a session will not typically be willing
   to wait until the dialog is established in order to begin compressing
   messages.  One of the biggest gains that SigComp can bring to SIP is
   the ability to compress the initial INVITE of a dialog, when the user
   is waiting for the session to be established.  Therefore, clients
   need a means to obtain a comp=sigcomp URI from their outbound proxy
   before the user decides to establish a session.

   One solution to this problem is manual configuration.  However,
   sometimes it is necessary to have clients configured in an automatic
   fashion.  Unfortunately, current mechanisms for SIP client
   configuration (e.g., using DHCP [6]) do not allow to provide the
   client with URI parameters.  In this case, the client SHOULD send an
   uncompressed OPTIONS request to its outbound proxy.  The outbound
   proxy can provide an alternative SIP URI with the comp=sigcomp
   parameter in a Contact header field in a 200 OK response to the
   OPTIONS.  The client can use this URI for subsequent requests that
   are sent through the same outbound proxy using compression.

   RFC 3261 [1] does not define how a proxy should respond to an OPTIONS
   request addressed to itself.  It only describes how servers respond
   to OPTIONS addressed to a particular user.  Section 11.2 of RFC 3261
   says:

      Contact header fields MAY be present in a 200 (OK) response and
      have the same semantics as in a 3xx response.  That is, they may
      list a set of alternative names and methods of reaching the user.

   We extend this behavior to proxy servers responding to OPTIONS
   addressed to them.  They MAY list a set of alternative URIs to
   contact the proxy.

   Note that receiving incoming requests (even initial INVITEs)
   compressed is not a problem, since user agents can REGISTER a SIP URI
   with comp=sigcomp in their registrar.  All incoming requests for the
   user will be sent to this SIP URI using compression.

5.   Sending a Response to a Client

   A response is sent to the host in the sent-by parameter of the Via
   header field.  If the topmost Via header field contains the parameter
   comp=sigcomp, the response SHOULD be compressed.  Otherwise, the
   response MUST NOT be compressed.

   In order to avoid asymmetric compression (i.e., two SIP entities
   exchanging compressed requests in one direction and uncompressed
   requests in the other direction) proxies need to rewrite their
   Record-Route entries in the responses.  A proxy performing Record-
   Route inspects the Record-Route header field in the response and the
   Contact header field in the request that triggered this response (see
   example in Section 9).  It looks for the URI of the next upstream
   (closer to the user agent client) hop in the route set.  If this URI
   contains the parameter comp=sigcomp, the proxy SHOULD add
   comp=sigcomp to its entry in the Record-Route header field.  If this
   URI does not contain the parameter comp=sigcomp, the proxy SHOULD
   remove comp=sigcomp (if it is present) from its entry in the Record-
   Route header field.

   The same way, a user agent server SHOULD add comp=sigcomp to the
   Contact header field of the response if the URI of the next upstream
   hop in the route set contained the parameter comp=sigcomp.

6.   Double Record-Routing

   Although proxies usually add zero or one Record-Route entries to a
   particular request, some proxies add two of them to avoid Record-
   Route rewriting.  A typical example of double Record-Routing is a SIP
   proxy that acts as a firewall between two networks.  Depending on
   which network a request comes from, it will be received on a
   different interface by the proxy.  The proxy adds one Record-Route
   entry for one interface and a second one for the other interface.
   This way, the proxy does not need to rewrite the Record-Route header
   field on the response.

   Proxies that receive compressed messages from one side of the dialog
   (e.g., upstream) and uncompressed messages from the other side (e.g.,
   downstream) MAY use the mechanism described above.

   If a proxy detects that the next-hop proxy for a request is the proxy
   itself and that the request will not be sent through the network, the
   proxy MAY choose not to compress the request even if the URI contains
   the comp=sigcomp parameter.

7.   Error Situations

   If a compressed SIP request arrives to a SIP server that does not
   understand SigComp, the server will not have any means to indicate
   the error to the client.  The message will be impossible to parse,
   and there will be no Via header field indicating an address to send
   an error response.

   If a SIP client sends a compressed request and the client transaction
   times out without having received any response, the client SHOULD
   retry the same request without using compression.  If the compressed
   request was sent over a TCP connection, the client SHOULD close that
   connection and open a new one to send the uncompressed request.
   Otherwise the server would not be able to detect the beginning of the
   new message.

8.   Augmented BNF

   This section provides the augmented Backus-Naur Form (BNF) of both
   parameters described above.

   The compression URI parameter is a "uri-parameter", as defined by the
   SIP ABNF (Section 25.1 of [1]):

      compression-param  =  "comp=" ("sigcomp" / other-compression)
      other-compression  =  token

   The Via compression parameter is a "via-extension", as defined by the
   SIP ABNF (Section 25.1 of [1]):

      via-compression    =  "comp" EQUAL ("sigcomp" / other-compression)
      other-compression  =  token

9.   Example

   The following example illustrates the use of the parameters defined
   above.  The call flow of Figure 1 shows an INVITE-200 OK-ACK
   handshake between a UAC and a UAS through two proxies.  Proxy P1 does
   not Record-Route but proxy P2 does.  Both proxies support
   compression, but they do not use it by default.

   UAC            P1            P2           UAS

    |(1)INVITE(c) |             |             |
    |------------>| (2) INVITE  |             |
    |             |------------>| (3) INVITE  |
    |             |             |------------>|
    |             |             | (4) 200 OK  |
    |             | (5) 200 OK  |<------------|
    |(6)200 OK(c) |<------------|             |
    |<------------|             |             |
    |             |  (7)ACK(c)  |             |
    |-------------------------->|   (8) ACK   |
    |             |             |------------>|
    |             |             |             |
    |             |             |             |

   Figure 1: INVITE transaction through two proxies

   Messages (1), (6) and (7) are compressed (c).

   We provide a partial description of the messages involved in this
   call flow below.  Only some parts of each message are shown, namely
   the Method name, the Request-URI and the Via, Route, Record-Route and
   Contact header fields.  We have not used a correct format for these
   header fields.  We have rather focus on the contents of the header
   fields and on the presence (or absence) of the "comp=sigcomp"
   parameter.

      (1) INVITE UAS
          Via: UAC;comp=sigcomp
          Route: P1;comp=sigcomp
          Contact: UAC;comp=sigcomp

   P1 is the outbound proxy of the UAC, and it supports SigComp.  The
   UAC is configured to send compressed traffic to P1, and therefore, it
   compresses the INVITE (1).  In addition, the UAC wants to receive
   future requests and responses for this dialog compressed.  Therefore,
   it adds the comp=Sigcomp parameter to the Via and to the Contact
   header fields.

      (2) INVITE UAS
          Via: P1
          Via: UAC;comp=sigcomp
          Route: P2
          Contact: UAC;comp=sigcomp

   P1 forwards the INVITE (2) to P2.  P1 does not use compression by
   default, so it sends the INVITE uncompressed to P2.

      (3) INVITE UAS
          Via: P2
          Via: P1
          Via: UAC;comp=sigcomp
          Record-Route: P2
          Contact: UAC;comp=sigcomp

   P2 forwards the INVITE (3) to the UAS.  P2 supports compression, but
   it does not use it by default.  Therefore, it sends the INVITE
   uncompressed.  P2 wishes to remain in the signalling path and
   therefore it Record-Routes.

      (4) 200 OK
          Via: P2
          Via: P1
          Via: UAC;comp=sigcomp
          Record-Route: P2
          Contact: UAS

   The UAS generates a 200 OK response and sends it to the host in the
   topmost Via, which is P2.

      (5) 200 OK
          Via: P1
          Via: UAC;comp=sigcomp
          Record-Route: P2;comp=sigcomp
          Contact: UAS

   P2 receives the 200 OK response.  P2 Record-Routed, so it inspects
   the Route set for this dialog.  For requests from the UAS towards the
   UAC (the opposite direction than the first INVITE), the next hop will
   be the Contact header field of the INVITE, because P1 did not
   Record-Route.  That Contact identified the UAC:

      Contact: UAC;comp=sigcomp

   Since the UAC wants to receive compressed requests (Contact of the
   INVITE), P2 assumes that the UAC would also like to send compressed
   requests (Record-Route of the 200 OK).  Therefore, P2 modifies its
   entry in the Record-Route header field of the 200 OK (5).  In the
   INVITE (3), P2 did not used the comp=sigcomp parameter.  Now it adds
   it in the 200 OK (5).  This will allow the UAC sending compressed
   requests within this dialog.

      (6) 200 OK
          Via: UAC;comp=sigcomp
          Record-Route: P2;comp=sigcomp
          Contact: UAS

   P1 sends the 200 OK (6) compressed to the UAC because the Via header
   field contained the comp=sigcomp parameter.

      (7) ACK UAS
          Via: UAC;comp=sigcomp
          Route: P2;comp=sigcomp
          Contact: UAC;comp=sigcomp

   The UAC sends the ACK (7) compressed directly to P2 (P1 did not
   Record-Route).

      (8) ACK UAS
          Via: P2
          Via: UAC;comp=sigcomp
          Contact: UAC;comp=sigcomp

   P2 sends the ACK (8) uncompressed to the UAS.

10.   Security Considerations

   A SIP entity receiving a compressed message has to decompress it and
   to parse it.  This requires slightly more processing power than only
   parsing a message.  This implies that a denial of service attack
   using compressed messages would be slightly worse than an attack with
   uncompressed messages.

   An attacker inserting the parameter comp=sigcomp in a SIP message
   could make a SIP entity send compressed messages to another SIP
   entity that did not support SigComp.  Appropriate integrity
   mechanisms should be used to avoid this attack.

11.   IANA Considerations

   This document defines the "comp" uri-parameter and via-extension.
   New values for "comp" are registered by the IANA at
   http://www.iana.org/assignments/sip-parameters when new signalling
   compression schemes are published in standards track RFCs.  The IANA
   Considerations section of the RFC MUST include the following
   information, which appears in the IANA registry along with the RFC
   number of the publication.

      o  Name of the compression scheme.

      o  Token value to be used. The token MAY be of any length, but
         SHOULD be no more than ten characters long.

   The only entry in the registry for the time being is:

   Compression scheme      Token      Reference
   ---------------------   ---------  ---------
   Signaling Compression   sigcomp    RFC 3486

12.   Acknowledgements

   Allison Mankin, Jonathan Rosenberg and Miguel Angel Garcia-Martin
   provided valuable comments on this memo.

13.   Normative References

   [1]  Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, A.,
        Peterson, J., Sparks, R., Handley, M. and E. Schooler, "SIP:
        Session Initiation Protocol", RFC 3261, June 2002.

   [2]  Price, R., Bormann, C., Christoffersson, J., Hannu, H., Liu, Z.
        and J. Rosenberg, "Signaling Compression (SigComp)", RFC 3320,
        January 2003.

   [3]  Bradner, S., "Key words for use in RFCs to indicate requirement
        levels", BCP 14, RFC 2119, March 1997.

14.   Informative References

   [4]  Mealling, M., "Dynamic Delegation Discovery System (DDDS) Part
        Three: The Domain Name System (DNS) Database", RFC 3403, October
        2002.

   [5]  Gulbrandsen, A., Vixie, P. and L. Esibov, "A DNS RR for
        specifying the location of services (DNS SRV)", RFC 2782,
        February 2000.

   [6]  Schulzrinne, H., "DHCP option for SIP servers", Work in
        Progress.

15.   Author's Address

   Gonzalo Camarillo
   Ericsson
   Advanced Signalling Research Lab.
   FIN-02420 Jorvas
   Finland

   EMail:  Gonzalo.Camarillo@ericsson.com

16.  Full Copyright Statement

   Copyright (C) The Internet Society (2003).  All Rights Reserved.

   This document and translations of it may be copied and furnished to
   others, and derivative works that comment on or otherwise explain it
   or assist in its implementation may be prepared, copied, published
   and distributed, in whole or in part, without restriction of any
   kind, provided that the above copyright notice and this paragraph are
   included on all such copies and derivative works.  However, this
   document itself may not be modified in any way, such as by removing
   the copyright notice or references to the Internet Society or other
   Internet organizations, except as needed for the purpose of
   developing Internet standards in which case the procedures for
   copyrights defined in the Internet Standards process must be
   followed, or as required to translate it into languages other than
   English.

   The limited permissions granted above are perpetual and will not be
   revoked by the Internet Society or its successors or assigns.

   This document and the information contained herein is provided on an
   "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
   TASK FORCE DISCLAIMS 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.

Acknowledgement

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