Network Working Group                                        H. Kitamura
Request for Comments: 3089                               NEC Corporation
Category: Informational                                       April 2001

A SOCKS-based IPv6/IPv4 Gateway Mechanism

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

Abstract

   This document describes a SOCKS-based IPv6/IPv4 gateway mechanism
   that enables smooth heterogeneous communications between the IPv6
   nodes and IPv4 nodes.

   It is based on the SOCKS protocol [SOCKSv5].  By applying the SOCKS
   mechanism to the heterogeneous communications and relaying two
   "terminated" IPv4 and IPv6 connections at the "application layer"
   (the SOCKS server), the SOCKS-based IPv6/IPv4 gateway mechanism is
   accomplished.

   Since it is accomplished without introducing new protocols, it
   provides the same communication environment that is provided by the
   SOCKS mechanism.  The same appearance is provided to the
   heterogeneous communications.  No conveniences or functionalities of
   current communications are sacrificed.

1. Introduction

   The SOCKS-based IPv6/IPv4 gateway mechanism is based on a mechanism
   that relays two "terminated" IPv4 and IPv6 connections at the
   "application layer" (the SOCKS server); its characteristics are
   inherited from those of the connection relay mechanism at the
   application layer and those of the native SOCKS mechanism.

2. Basic SOCKS-based Gateway Mechanism

   Figure 1 shows the basic SOCKS-based gateway mechanism.

                  Client C       Gateway G     Destination D
               +-----------+     (Server)
               |Application|
           +-->+===========+  +-------------+  +-----------+
      same-+   |*SOCKS Lib*|  |  *Gateway*  |  |Application|
       API +-->+===========+  +=====---=====+  +-----------+
               | Socket DNS|  | Socket  DNS |  | Socket DNS|
               +-----------+  +-------------+  +-----------+
               | [ IPv X ] |  |[IPvX]|(IPvY)|  | ( IPv Y ) |
               +-----------+  +-------------+  +-----------+
               |Network I/F|  | Network I/F |  |Network I/F|
               +-----+-----+  +---+-----+---+  +-----+-----+
                     |            |     |            |
                     +============+     +------------+
                       socksified           normal
                       connection         connection
                      (ctrl)+data          data only

                Fig. 1 Basic SOCKS-based Gateway Mechanism

   In this figure, the Client C initiates the communication to the
   Destination D.  Two new functional blocks are introduced and they
   compose the mechanism.

   One, *Socks Lib*, is introduced into the client side (Client C) (this
   procedure is called "socksifying").  The *Socks Lib* is located
   between the application layer and the socket layer, and can replace
   applications' socket APIs and DNS name resolving APIs (e.g.,
   gethostbyname(), getaddrinfo() etc.).  There is a mapping table in it
   for a "DNS name resolving delegation" feature (described below).
   Each socksified application has its own *Socks Lib*.

   The other, *Gateway*, is installed on the IPv6 and IPv4 dual stack
   node (Gateway G).  It is an enhanced SOCKS server that enables any
   types of protocol combination relays between Client C (IPvX) and
   Destination D (IPvY).  When the *Socks Lib* invokes a relay, one
   corresponding *Gateway* process (thread) is spawned from the parent
   *Gateway* to take charge of the relay connection.

   The following four types of combinations of IPvX and IPvY are
   possible in the mechanism.

    type C ------ G ------ D
           [IPvX]   (IPvY)
     A      IPv4     IPv4       homogeneous (normal SOCKS)
     B      IPv4     IPv6     * heterogeneous *
     C      IPv6     IPv4     * heterogeneous *
     D      IPv6     IPv6       homogeneous

   Type A is supported by the normal SOCKS mechanism.  Type B and C are
   the main targets for the SOCKS-based IPv6/IPv4 gateway mechanism.
   They provide heterogeneous communications.  Type D can be supported
   by the natural extension of the SOCKS mechanism, because it is a
   homogeneous communication.

   Since the *Socks Lib* communicates with the *Gateway* by using SOCKS
   protocol [SOCKSv5], the connection between them (the Client C and the
   Gateway G) is a special connection and is called a "socksified
   connection".  It can transfer not only data but also control
   information (e.g., the location information of Destination D).

   The connection between the Gateway G and the Destination D is a
   normal connection.  It is not modified (socksified).  A server
   application that runs on Destination D does not notice the existence
   of the Client C.  It recognizes that the peer node of the connection
   is the Gateway G (not Client C).

   No new protocols are introduced to the SOCKS protocol [SOCKSv5] to
   accomplish the mechanism.

   * Packet Size Adjustment

     Since the length of the IPv6 header is different from that of the
     IPv4 header, it is necessary to consider the packet size adjustment
     in heterogeneous communications.  If this is not taken into
     consideration, the packet size may exceed the MTU of the network.

     In the SOCKS-based IPv6/IPv4 gateway mechanism, it never exceeds
     the MTU, because the mechanism is based on relaying two
     "terminated" connections at the "application layer".  The relayed
     data is a simple data stream for the application, and the packet
     size is naturally adjusted at each relayed connection side.

   * Authenticated Relay

     Since the SOCKS is originally designed for firewall systems and it
     has various authentication methods, the relayed connections can be
     authenticated by the native SOCKS authentication methods.

3. DNS Name Resolving Procedure

   In all communication applications, it is a necessary to obtain
   destination IP address information to start a communication.  It is,
   however, theoretically impossible for the heterogeneous
   communications to obtain correct information, because an existing
   IPv4 application can not deal with an IPv6 address.  It prepares only
   a 4-byte address space to store an IP address information, and it can
   not store an IPv6 address information into there.  This is a critical
   problem caused by differences in address length.

   In order to solve the problem, a feature called "DNS name resolving
   delegation" is used in the SOCKS-based IPv6/IPv4 gateway mechanism.
   The feature involves the delegating of DNS name resolving actions at
   the source node (Client C) to the relay server (Gateway G).  Since
   the relay server is an IPv4 and IPv6 dual stack node, DNS name
   resolving queries for any address family types of destinations can be
   made without causing any problems.  Therefore, it is not necessary to
   modify the existing DNS mechanism at all.

   The feature supports not only the case in which a destination logical
   host name (FQDN) information is given but also the case in which a
   destination literal (numerical) IP address is given.  The latter case
   is supported in almost the same way as the former case.  Since the
   literal IPv6 address expression includes colons (":"), it is
   identified as an FQDN (not a literal IPv4 address) for the IPv4
   application.

   The SOCKS protocol specification [SOCKSv5] defines that IPv4 address,
   IPv6 address, and DOMAINNAME (FQDN) information can be used in the
   ATYP (address type) field of the SOCKS protocol format.  In the "DNS
   name resolving delegation" feature, the DOMAINNAME (FQDN) information
   is used in the ATYP (address type) field.  The FQDN information is
   transferred from the Client C to the Gateway G to indicate the
   Destination D.

   In order to solve the formerly explained critical problem, an
   appropriate "fake IP" address is introduced in the feature, and it is
   used as a virtual destination IP address for a socksified
   application.  A mapping table is also introduced in the *Socks Lib*
   (at the Client C) to manage mappings between "fake IP" and "FQDN".  A
   "fake IP" address is used as a key to look up the corresponding
   "FQDN" information.  The mapping table is local and independent of
   other applications or their *Socks Lib*s.

   The transparentness to applications is maintained in the feature.
   Nothing special is required to execute it except socksifying the
   applications.  Since DNS name resolving APIs are replaced by the
   *Socks Lib*, the "DNS name resolving delegation" is executed
   internally merely by calling the DNS name resolving APIs in ordinal
   methods.

   The "DNS name resolving delegation" is accomplished only when FQDN
   information is used in the ATYP (address type) field of the SOCKS
   command.  Therefore, it is mandatory to do so for heterogeneous
   communications.  The method of using FQDN information in the ATYP
   field depends on the configuration setting and implementation of the
   SOCKS protocol.  In order to simplify the discussion, only the case
   in which the FQDN information is used in the ATYP field is discussed
   here.

   The detailed internal procedure of the "DNS name resolving
   delegation" and address mapping management related issues are
   described as follows.

   1. An application on the source node (Client C) tries to get the
      IP address information of the destination node (Destination D) by
      calling the DNS name resolving function (e.g., gethostbyname()).
      At this time, the logical host name ("FQDN") information of the
      Destination D is passed to the application's *Socks Lib* as an
      argument of called APIs.

   2. Since the *Socks Lib* has replaced such DNS name resolving APIs,
      the real DNS name resolving APIs is not called here.  The argued
      "FQDN" information is merely registered into a mapping table in
      *Socks Lib*, and a "fake IP" address is selected as information
      that is replied to the application from a reserved special IP
      address space that is never used in real communications (e.g.,
      0.0.0.x).  The address family type of the "fake IP" address must be
      suitable for requests called by the applications.  Namely, it must
      belong to the same address family of the Client C, even if the
      address family of the Destination D is different from it.  After
      the selected "fake IP" address is registered into the mapping
      table as a pair with the "FQDN", it is replied to the application.

   3. The application receives the "fake IP" address, and prepares a
      "socket".  The "fake IP" address information is used as an element
      of the "socket".  The application calls socket APIs (e.g.,
      connect()) to start a communication.  The "socket" is used as an
      argument of the APIs.

   4. Since the *Socks Lib* has replaced such socket APIs, the real
      socket function is not called.  The IP address information of the
      argued socket is checked.  If the address belongs to the special
      address space for the fake address, the matched registered "FQDN"
      information of the "fake IP" address is obtained from the mapping
      table.

   5. The "FQDN" information is transferred to the *Gateway* on the
      relay server (Gateway G) by using the SOCKS command that is
      matched to the called socket APIs.  (e.g., for connect(), the
      CONNECT command is used.)

   6. Finally, the real DNS name resolving API (e.g., getaddrinfo()) is
      called at the *Gateway*.  At this time, the received "FQDN"
      information via the SOCKS protocol is used as an argument of the
      called APIs.

   7. The *Gateway* obtains the "real IP" address from a DNS server,
      and creates a "socket".  The "real IP" address information is used
      as an element of the "socket".

   8. The Gateway calls socket APIs (e.g., connect()) to communicate
      with the Destination D.  The "socket" is used as an argument of the
      APIs.

   The problem with the feature is that failures of the DNS name
   resolving process are detected incorrectly at the source node (Client
   C).  They are detected as connection-establishment failures.

   (Restrictions on applicability of "fake IP" address, etc., are
   described in Section 5.)

   * Operations for Address Management (reservation, mapping etc.)

   The SOCKS-based gateway mechanism does not require the reserving of a
   wide global address space for the address mapping, and complex
   address allocation and garbage-collection mechanisms are not
   necessary.

   Such address management operations are done at the *Socks Lib* by
   using the fake IP address and the mapping table for the DNS name
   resolving delegation.  Since the mapping table is prepared in each
   application, it is locally closed and independent of other
   applications.  Therefore, it is easy to manage the table, and it is
   not necessary to reserve a wide global address space.

4. Multiple Chained Relay Mechanism (Advanced usage)

   The SOCKS-based gateway mechanism has the flexibility to support
   multiple chained relay topologies.  With the mechanism, IPv4 and IPv6
   mixed various communication topologies are accomplished.

   Figure 2 shows the structure of the multiple chained relay mechanism.

        Client C       Gateway G1       Gateway G2    Destination D
     +-----------+     (Server 1)       (Server 2)
     |Application|
     +===========+  +-------------+  +-------------+  +-----------+
     |*SOCKS Lib*|  |  *Gateway1* |  |  *Gateway2* |  |Application|
     +===========+  +=====---=====+  +=====---=====+  +-----------+
     | Socket DNS|  | Socket  DNS |  | Socket  DNS |  | Socket DNS|
     +-----------+  +-------------+  +-------------+  +-----------+
     | [ IPv X ] |  |[IPvX]|(IPvY)|  |(IPvY)|{IPvZ}|  | { IPv Z } |
     +-----------+  +-------------+  +-------------+  +-----------+
     |Network I/F|  | Network I/F |  | Network I/F |  |Network I/F|
     +-----+-----+  +---+-----+---+  +---+-----+---+  +-----+-----+
           |            |     |          |     |            |
           +============+     +==========+     +------------+
             socksified        socksified          normal
             connection        connection        connection
            (ctrl)+data       (ctrl)+data         data only

                  Fig. 2 Multiple Chained Relay Mechanism

   In this figure, the source node (Client C) initiates the
   communication with the destination (Destination D).  Underneath, the
   connection is replaced with three connections, and they are relayed
   at the two relay servers (Gateway G1 and G2).  The *Gateway* includes
   the same type of functions of *Socks Lib*.  By enabling the *Socks
   Lib* functions at the *Gateway1* on the first relay server (Gateway
   G1), the multiple chained relay topology is accomplished.

   There is no limitation on the number of relay operations between the
   source node and the final destination node.  It is possible to have
   more than two intermediate relay servers.  To simplify the
   explanation, a twice-relayed topology is shown here.

   Since the multiple chained relay is more complex than one-time relay
   and causes complexity, it is recommended that the multiple chained
   relay communication should be used only when it is necessary for some
   reason (e.g., usable protocols or topologies are limited by routers
   etc.).

5. Applicability statement

   The SOCKS-based gateway mechanism requests socksification of
   applications (install *Socks Lib*) to accomplish heterogeneous
   communications.  It is not necessary to modify (change source codes
   and recompile them, etc.) the applications, because typical
   socksification is done by changing the linking order of dynamic link
   libraries (specifically, by linking the SOCKS dynamic link library
   before the dynamic link libraries for normal socket and DNS name
   resolving APIs).

   The mechanism does not request modification of the DNS system,
   because the DNS name resolving procedure at the Client C is delegated
   to the dual stack node Gateway G.

   Other than the socksification, the SOCKS-based gateway mechanism has
   the following three types of constraints.

   1. Essential constraints:

      Constraints are caused by the address length difference between
      IPv4 and IPv6.

      Functions that request an IP address as one of the return values
      (e.g., getpeername() and getsockname() etc.) can not provide the
      correct IP address as a return value.  However, a suitable port
      value can be provided, because IPv4 and IPv6 use the same size
      port space and an appropriate port information is transferred by
      the SOCKS protocol.

   2. Constraints of the SOCKS mechanism:

      Since the current SOCKS system can not socksify all of the tricky
      applications in which extraordinary manners are used to create
      connections, the SOCKS-based gateway mechanism can not be applied
      to them.

   3. Constraints to deal with the fake address:

      The fake address must be dealt with as a temporary value at the
      application.  It is used as a key value in the mapping table for
      the "DNS name resolving delegation" feature.  When the application
      is finished and the mapping table disappears, the fake address
      information must be also released.

      Even if it is recorded permanently (e.g., recorded as a bookmark),
      serious problems will not occur.  The recorded fake address
      information will merely become useless, because fake address

      information is taken from a reserved special IP address space that
      is never used in real communications (e.g., 0.0.0.x) and such a
      information is useless for the normal communication applications.
      Furthermore, such cases will be rare because most applications
      usually record FQDN information (not fake IP address information)
      to the bookmark, etc.

5.1 Native SOCKS mechanism considerations

   The characteristics of the SOCKS-based IPv6/IPv4 gateway mechanism
   are inherited from those of the native SOCKS mechanism.  Therefore,
   consideration issues of the native SOCKS mechanism are discussed in
   this section.

   The SOCKSv5 protocol is composed of three commands (CONNECT, BIND and
   UDP ASSOCIATE).  All of three commands can be applied in the SOCKS-
   based IPv6/IPv4 gateway mechanism.

   This document is described with assuming the usage of the CONNECT
   command mainly, because the CONNECT command is the main and most
   frequently used command in the SOCKS mechanism.  Since the CONNECT
   command does not have clear week points, we can use it freely without
   considerations.

   The other (BIND and UDP ASSOCIATE) commands have the following weak
   points.  So, we have to consider these points when we use the BIND or
   UDP ASSOCIATE commands in the mechanism.

   The BIND command is basically designed to support reverse-channel
   rendezvous of the FTP type applications.  So, general usages of the
   BIND command may cause problems.

   The UDP ASSOCIATE command is basically designed for simple UDP
   applications (e.g., archie).  It is not general enough to support a
   large class of applications that use both TCP and UDP.

6. Security Considerations

   Since the SOCKS-based IPv6/IPv4 gateway mechanism is based on SOCKSv5
   protocol, the security feature of the mechanism matches that of
   SOCKSv5.  It is described in the Security Considerations section of
   the SOCKS Protocol Version 5 [SOCKSv5].

   The mechanism is based on relaying two "terminated" connections at
   the "application layer".  The end-to-end security is maintained at
   each of the relayed connections (i.e., between Client C and Gateway
   G, and between Gateway G and Destination D).  The mechanism does not
   provide total end-to-end security relay between the original source
   (Client C) and the final destination (Destination D).

Appendix A. Implementations

   Currently, there are two independent implementations of the SOCKS-
   based IPv6/IPv4 gateway mechanism.  Both of them are open to the
   public.

   One is NEC's implementation.  Its source codes are available at the
   following URL.

            http://www.socks.nec.com/

   The other is Fujitsu Lab.'s implementation, which is called
   "SOCKS64".  Its source codes are available at the following URL.

            ftp://ftp.kame.net/pub/kame/misc/socks64-...

References

   [SOCKSv5]    Leech, M., Ganis, M., Lee, Y., Kuris, R., Koblas, D. and
                L. Jones, "SOCKS Protocol V5", RFC 1928, April 1996.

   [TRANSMECH]  Gilligan, R. and E. Nordmark, "Transition Mechanisms for
                IPv6 Hosts and Routers", RFC 2893, August 2000.

   [IPv6]       Deering, S. and R. Hinden, "Internet Protocol, Version 6
                (IPv6) Specification", RFC 2460, December 1998.

   [INET99]     H. Kitamura, "Entering the IPv6 communication world by
                the SOCKS-based IPv6/IPv4 Translator", in Proceedings of
                INET99, July 1999.

Author's Address

   Hiroshi Kitamura
   NEC Corporation
   Development Laboratories
   (Igarashi Building 4F) 11-5, Shibaura 2-Chome,
   Minato-Ku, Tokyo 108-8557, JAPAN

   Phone: +81 (3) 5476-1071
   Fax:   +81 (3) 5476-1005
   EMail: kitamura@da.jp.nec.com

Full Copyright Statement

   Copyright (C) The Internet Society (2001).  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.