Network Working Group A. Durand
Request for Comments: 3053 SUN Microsystems, Inc
Category: Informational P. Fasano
I. Guardini
CSELT S.p.A.
D. Lento
TIM
January 2001

IPv6 Tunnel Broker

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 Internet Society (2001). All Rights Reserved.

Abstract

The IPv6 global Internet as of today uses a lot of tunnels over the
existing IPv4 infrastrUCture. Those tunnels are difficult to
configure and maintain in a large scale environment. The 6bone has
proven that large sites and Internet Service Providers (ISPs) can do
it, but this process is too complex for the isolated end user who
already has an IPv4 connection and would like to enter the IPv6
world. The motivation for the development of the tunnel broker model
is to help early IPv6 adopters to hook up to an existing IPv6 network
(e.g., the 6bone) and to get stable, permanent IPv6 addresses and DNS
names. The concept of the tunnel broker was first presented at
Orlando's IETF in December 1998. Two implementations were
demonstrated during the Grenoble IPng & NGtrans interim meeting in
February 1999.

1. Introduction

The growth of IPv6 networks started mainly using the transport
facilities offered by the current Internet. This led to the
development of several techniques to manage IPv6 over IPv4 tunnels.
At present most of the 6bone network is built using manually
configured tunnels over the Internet. The main drawback of this
approach is the overwhelming management load for network
administrators, who have to perform extensive manual configuration
for each tunnel. Several attempts to reduce this management overhead

have already been proposed and each of them presents interesting
advantages but also solves different problems than the Tunnel Broker,
or poses drawbacks not present in the Tunnel Broker:

- the use of automatic tunnels with IPv4 compatible addresses [1]
is a simple mechanism to establish early IPv6 connectivity
among isolated dual-stack hosts and/or routers. The problem
with this approach is that it does not solve the address
exhaustion problem of IPv4. Also there is a great fear to
include the complete IPv4 routing table into the IPv6 world
because this would worsen the routing table size problem
multiplying it by 5;

- 6over4 [2] is a site local transition mechanism based on the
use of IPv4 multicast as a virtual link layer. It does not
solve the problem of connecting an isolated user to the global
IPv6 Internet;

- 6to4 [3] has been designed to allow isolated IPv6 domains,
attached to a wide area network with no native IPv6 support
(e.g., the IPv4 Internet), to communicate with other such IPv6
domains with minimal manual configuration. The idea is to
embed IPv4 tunnel addresses into the IPv6 prefixes so that any
domain border router can automatically discover tunnel
endpoints for outbound IPv6 traffic.

The Tunnel Broker idea is an alternative approach based on the
provision of dedicated servers, called Tunnel Brokers, to
automatically manage tunnel requests coming from the users. This
approach is eXPected to be useful to stimulate the growth of IPv6
interconnected hosts and to allow early IPv6 network providers to
provide easy Access to their IPv6 networks.

The main difference between the Tunnel Broker and the 6to4 mechanisms
is that the they serve a different segment of the IPv6 community:

- the Tunnel Broker fits well for small isolated IPv6 sites, and
especially isolated IPv6 hosts on the IPv4 Internet, that want
to easily connect to an existing IPv6 network;

- the 6to4 approach has been designed to allow isolated IPv6
sites to easily connect together without having to wait for
their IPv4 ISPs to deliver native IPv6 services. This is very
well suited for extranet and virtual private networks. Using
6to4 relays, 6to4 sites can also reach sites on the IPv6
Internet.

In addition, the Tunnel Broker approach allows IPv6 ISPs to easily
perform access control on the users enforcing their own policies on
network resources utilization.

This document is intended to present a framework describing the
guidelines for the provision of a Tunnel Broker service within the
Internet. It does not specify any protocol but details the general
architecture of the proposed approach. It also outlines a set of
viable alternatives for implementing it. Section 2 provides an
overall description of the Tunnel Broker model; Section 3 reports
known limitations to the model; Section 4 briefly outlines other
possible applications of the Tunnel Broker approach; Section 5
addresses security issues.

2. Tunnel Broker Model

Tunnel brokers can be seen as virtual IPv6 ISPs, providing IPv6
connectivity to users already connected to the IPv4 Internet. In the
emerging IPv6 Internet it is expected that many tunnel brokers will
be available so that the user will just have to pick one. The list
of the tunnel brokers should be referenced on a "well known" web page
(e.g. on http://www.ipv6.org) to allow users to choose the "closest"
one, the "cheapest" one, or any other one.

The tunnel broker model is based on the set of functional elements
depicted in figure 1.

+------+
/tunnel
/ server
/
/ +------+
+----------+ +------+/ +------+
dual-stack tunnel tunnel
node <--->broker<--->server
(user)
+----------+ +------+ +------+
+------+
tunnel end-point v tunnel
/ +---+ server
DNS
+---+ +------+

tunnel end-point
/

+---------------------------+
+-----------------------------+
IPv6 over IPv4 tunnel

Figure 1: the Tunnel Broker model

2.1 Tunnel Broker (TB)

The TB is the place where the user connects to register and activate
tunnels. The TB manages tunnel creation, modification and deletion
on behalf of the user.

For scalability reasons the tunnel broker can share the load of
network side tunnel end-points among several tunnel servers. It
sends configuration orders to the relevant tunnel server whenever a
tunnel has to be created, modified or deleted. The TB may also
register the user IPv6 address and name in the DNS.

A TB must be IPv4 addressable. It may also be IPv6 addressable, but
this is not mandatory. Communications between the broker and the
servers can take place either with IPv4 or IPv6.

2.2 Tunnel server (TS)

A TS is a dual-stack (IPv4 & IPv6) router connected to the global
Internet. Upon receipt of a configuration order coming from the TB,
it creates, modifies or deletes the server side of each tunnel. It
may also maintain usage statistics for every active tunnel.

2.3 Using the Tunnel Broker

The client of the Tunnel Broker service is a dual-stack IPv6 node
(host or router) connected to the IPv4 Internet. Approaching the TB,
the client should be asked first of all to provide its identity and
credentials so that proper user authentication, authorization and
(optionally) accounting can be carried out (e.g., relying on existing
AAA facilities such as RADIUS). This means that the client and the
TB have to share a pre-configured or automatically established
security association to be used to prevent unauthorized use of the
service. With this respect the TB can be seen as an access-control
server for IPv4 interconnected IPv6 users.

Once the client has been authorized to access the service, it should
provide at least the following information:

- the IPv4 address of the client side of the tunnel;

- a name to be used for the registration in the DNS of the global
IPv6 address assigned to the client side of the tunnel;

- the client function (i.e., standalone host or router).

Moreover, if the client machine is an IPv6 router willing to provide
connectivity to several IPv6 hosts, the client should be asked also
to provide some information about the amount of IPv6 addresses
required. This allows the TB to allocate the client an IPv6 prefix
that fits its needs instead of a single IPv6 address.

The TB manages the client requests as follows:

- it first designates (e.g., according to some load sharing
criteria defined by the TB administrator) a Tunnel Server to be
used as the actual tunnel end-point at the network side;

- it chooses the IPv6 prefix to be allocated to the client; the
prefix length can be anything between 0 and 128, most common
values being 48 (site prefix), 64 (subnet prefix) or 128 (host
prefix);

- it fixes a lifetime for the tunnel;

- it automatically registers in the DNS the global IPv6 addresses
assigned to the tunnel end-points;

- it configures the server side of the tunnel;

- it notifies the relevant configuration information to the
client, including tunnel parameters and DNS names.

After the above configuration steps have been carried out (including
the configuration of the client), the IPv6 over IPv4 tunnel between
the client host/router and the selected TS is up and working, thus
allowing the tunnel broker user to get access to the 6bone or any
other IPv6 network the TS is connected to.

2.4 IPv6 address assignment

The IPv6 addresses assigned to both sides of each tunnel must be
global IPv6 addresses belonging to the IPv6 addressing space managed
by the TB.

The lifetime of these IPv6 addresses should be relatively long and
potentially longer than the lifetime of the IPv4 connection of the
user. This is to allow the client to get semipermanent IPv6
addresses and associated DNS names even though it is connected to the
Internet via a dial-up link and gets dynamically assigned IPv4
addresses through DHCP.

2.5 Tunnel management

Active tunnels consume precious resources on the tunnel servers in
terms of memory and processing time. For this reason it is advisable
to keep the number of unused tunnels as small as possible deploying a
well designed tunnel management mechanism.

Each IPv6 over IPv4 tunnel created by the TB should at least be
assigned a lifetime and removed after its expiration unless an
explicit lifetime extension request is submitted by the client.

Obviously this is not an optimal solution especially for users
accessing the Internet through short-lived and dynamically addressed
IPv4 connections (e.g., dial-up links). In this case a newly
established tunnel is likely to be used just for a short time and
then never again, in that every time the user reconnects he gets a
new IPv4 address and is therefore obliged either to set-up a new
tunnel or to update the configuration of the previous one. In such a
situation a more effective tunnel management may be achieved by
having the TS periodically deliver to the TB IPv6 traffic and
reachability statistics for every active tunnel. In this way, the TB
can enforce a tunnel deletion after a period of inactivity without
waiting for the expiration of the related lifetime which can be
relatively longer (e.g., several days).

Another solution may be to implement some kind of tunnel management
protocol or keep-alive mechanism between the client and the TS (or
between the client and the TB) so that each tunnel can be immediately
released after the user disconnects (e.g., removing his tunnel end-
point or tearing down his IPv4 connection to the Internet). The
drawback of this policy mechanism is that it also requires a software
upgrade on the client machine in order to add support for the ad-hoc
keep-alive mechanism described above.

Moreover, keeping track of the tunnel configuration even after the
user has disconnected from the IPv4 Internet may be worth the extra
cost. In this way, in fact, when the user reconnects to the
Internet, possibly using a different IPv4 address, he could just
restart the tunnel by getting in touch with the TB again. The TB
could then order a TS to re-create the tunnel using the new IPv4
address of the client but reusing the previously allocated IPv6
addresses. That way, the client could preserve a nearly permanent
(static) IPv6 address even though its IPv4 address is dynamic. It
could also preserve the associated DNS name.

2.6 Interactions between client, TB, TS and DNS

As previously stated, the definition of a specific set of protocols
and procedures to be used for the communication among the various
entities in the Tunnel Broker architecture is outside of the scope of
the present framework document. Nevertheless, in the reminder of
this section some viable technical alternatives to support client-TB,
TB-TS and TB-DNS interactions are briefly described in order to help
future implementation efforts or standardization initiatives.

The interaction between the TB and the user could be based on http.
For example the user could provide the relevant configuration
information (i.e., the IPv4 address of the client side of the tunnel,
etc.) by just filling up some forms on a Web server running on the
TB. As a result the server could respond with an Html page stating
that the server end-point of the tunnel is configured and displaying
all the relevant tunnel information.

After that, the most trivial approach would be to leave the user to
configure the client end-point of the tunnel on his own. However, it
should be highly valuable to support a mechanism to automate this
procedure as much as possible.

Several options may be envisaged to assist the Tunnel Broker user in
the configuration of his dual-stack equipment. The simplest option
is that the TB could just prepare personalized activation and de-
activation scripts to be run off-line on the client machine to
achieve easy set-up of the client side tunnel end-point. This

solution is clearly the easiest to implement and operate in that it
does not require any software extension on the client machine.
However, it raises several security concerns because it may be
difficult for the user to verify that previously downloaded scripts
do not perform illegal or dangerous operations once executed.

The above described security issues could be elegantly overcome by
defining a new MIME (Multipurpose Internet Mail Extension) content-
type (e.g., application/tunnel) [4,5] to be used by the TB to deliver
the tunnel parameters to the client. In this case, there must be a
dedicated agent running on the client to process this information and
actually set-up the tunnel end-point on behalf of the user. This is
a very attractive approach which is worth envisaging. In particular,
the definition of the new content-type might be the subject of a
future ad-hoc document.

Several options are available also to achieve proper interaction
between the broker and the Tunnel Servers. For example a set of
simple RSH commands over IPsec could be used for this purpose.
Another alternative could be to use SNMP or to adopt any other
network management solution.

Finally, the Dynamic DNS Update protocol [6] should be used for
automatic DNS update (i.e., to add or delete AAAA, A6 and PTR records
from the DNS zone reserved for Tunnel Broker users) controlled by the
TB. A simple alternative would be for the TB to use a small set of
RSH commands to dynamically update the direct and inverse databases
on the authoritative DNS server for the Tunnel Broker users zone
(e.g. broker.isp-name.com).

2.7 Open issues

Real usage of the TB service may require the introduction of
accounting/billing functions.

3. Known limitations

This mechanism may not work if the user is using private IPv4
addresses behind a NAT box.

4. Use of the tunnel broker concept in other areas

The Tunnel Broker approach might be efficiently exploited also to
automatically set-up and manage any other kind of tunnel, such as a
multicast tunnel (e.g., used to interconnect multicast islands within
the unicast Internet) or an IPsec tunnel.

Moreover, the idea of deploying a dedicated access-control server,
like the TB, to securely authorize and assist users that want to gain
access to an IPv6 network might prove useful also to enhance other
transition mechanisms. For example it would be possible to exploit a
similar approach within the 6to4 model to achieve easy relay
discovery. This would make life easier for early 6to4 adopters but
would also allow the ISPs to better control the usage of their 6to4
relay facilities (e.g., setting up appropriate usage policies).

5. Security Considerations

All the interactions between the functional elements of the proposed
architecture need to be secured:

- the interaction between the client and TB;

- the interaction between the TB and the Tunnel Server;

- the interaction between the TB and the DNS.

The security techniques adopted for each of the required interactions
is dependent on the implementation choices.

For the client-TB interaction, the usage of http allows the
exploitation of widely adopted security features, such as SSL (Secure
Socket Layer) [7], to encrypt data sent to and downloaded from the
web server. This also makes it possible to rely on a simple
"username" and "passWord" authentication procedure and on existing
AAA facilities (e.g., RADIUS) to enforce access-control.

For the TB-TS interaction secure SNMP could be adopted [8,9,10]. If
the dynamic DNS update procedure is used for the TB-DNS interaction,
the security issues are the same as discussed in [11]. Otherwise, if
a simpler approach based on RSH commands is used, standard IPsec
mechanisms can be applied [12].

Furthermore, if the configuration of the client is achieved running
scripts provided by the TB, these scripts must be executed with
enough privileges to manage network interfaces, such as an
administrator/root role. This can be dangerous and should be
considered only for early implementations of the Tunnel Broker
approach. Transferring tunnel configuration parameters in a MIME
type over https is a more secure approach.

In addition a loss of confidentiality may occur whenever a dial-up
user disconnects from the Internet without tearing down the tunnel
previously established through the TB. In fact, the TS keeps
tunneling the IPv6 traffic addressed to that user to his old IPv4

address regardless of the fact that in the meanwhile that IPv4
address could have been dynamically assigned to another subscriber of
the same dial-up ISP. This problem could be solved by implementing
on every tunnel the keep-alive mechanism outlined in section 2.5 thus
allowing the TB to immediately stop IPv6 traffic forwarding towards
disconnected users.

Finally TBs must implement protections against denial of service
attacks which may occur whenever a malicious user exhausts all the
resources available on the tunnels server by aSKINg for a lot of
tunnels to be established altogether. A possible protection against
this attack could be achieved by administratively limiting the number
of tunnels that a single user is allowed to set-up at the same time.

6. Acknowledgments

Some of the ideas refining the tunnel broker model came from
discussion with Perry Metzger and Marc Blanchet.

7. References

[1] Gilligan, R. and E. Nordmark, "Transition Mechanisms for IPv6
Hosts and Routers", RFC1933, April 1996.

[2] Carpenter, B. and C. Jung, "Transmission of IPv6 over IPv4
Domains without Explicit Tunnels", RFC2529, March 1999.

[3] Carpenter, B. and K. Moore, "Connection of IPv6 Domains via IPv4
Clouds without Explicit Tunnels", Work in Progress.

[4] Freed, N. and N. Borenstein, "Multipurpose Internet Mail
Extensions (MIME) Part One: Format of Internet Message Bodies,
RFC2045, November 1996.

[5] Freed, N. and N. Borenstein, "Multipurpose Internet Mail
Extensions (MIME) Part Two: Media Types", RFC2046, November
1996.

[6] Vixie, P., Editor, Thomson, T., Rekhter, Y. and J. Bound,
"Dynamic Updates in the Domain Name System (DNS UPDATE)", RFC
2136, April 1997.

[7] Guttman, E., Leong, L. and G. Malkin, "Users' Security
Handbook", FYI 34, RFC2504, February 1999.

[8] Wijnen, B., Harrington, D. and R. Presuhn, "An Architecture for
Describing SNMP Management Frameworks", RFC2571, April 1999.

[9] Blumenthal, U. and B. Wijnen, "User-based Security Model (USM)
for version 3 of the Simple Network Management Protocol
(SNMPv3)", RFC2574, April 1999.

[10] Wijnen, B., Presuhn, R. and K. McCloghrie, "View-based Access
Control Model (VACM) for the Simple Network Management Protocol
(SNMP)", RFC2575, April 1999.

[11] Eastlake, D., "Secure Domain Name System Dynamic Update", RFC
2137, April 1997.

[12] Kent, S. and R. Atkinson, "Security Architecture for the
Internet Protocol", RFC2401, November 1998.

8. Authors' Addresses

Alain Durand
SUN Microsystems, Inc
901 San Antonio Road
MPK17-202
Palo Alto, CA 94303-4900
USA

Phone: +1 650 786 7503
EMail: Alain.Durand@sun.com

Paolo Fasano S.p.A.
CSELT S.p.A.
Switching and Network Services - Networking
via G. Reiss Romoli, 274
10148 TORINO
Italy

Phone: +39 011 2285071
EMail: paolo.fasano@cselt.it

Ivano Guardini
CSELT S.p.A.
Switching and Network Services - Networking
via G. Reiss Romoli, 274
10148 TORINO
Italy

Phone: +39 011 2285424
EMail: ivano.guardini@cselt.it

Domenico Lento
TIM
Business Unit Project Management
via Orsini, 9
90100 Palermo
Italy

Phone: +39 091 7583243
EMail: dlento@mail.tim.it

9. 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 RFCEditor function is currently provided by the
Internet Society.