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

Versions: 00 01 02

INTERNET DRAFT                                              C. Huitema
<draft-ietf-midcom-scenarios-02.txt>                         Microsoft
Expires November 17, 2001                                 May 17, 2001

                             MIDCOM Scenarios

Status of this memo

This document is an Internet-Draft and is in full conformance with
all provisions of Section 10 of RFC2026.

This document is an Internet-Draft. 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.

Internet-Drafts are draft documents valid for a maximum of six
months and may be updated, replaced, or obsoleted by other documents
at any time.  It is inappropriate to use Internet- Drafts as
reference material or to cite them other than as "work in progress."

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

As trusted third parties are increasingly being asked to make policy
decisions on behalf of the various entities participating in an
application's operation, a need has developed for applications to be
able to communicate their needs to the devices in the network that
provide transport policy enforcement. Examples of these devices
include firewalls, network address translators (both within and
between address families), signature management for intrusion
detection systems, and multimedia buffer management. These devices
are a subset of what can be referred to as 'middle boxes.' This
document describes traversal scenarios that a 'middle box traversal
protocol' should enable.

1       Introduction

In order to delineate the requirement of the MIDCOM protocol, we
present here a set of scenarios that should be enabled by this
protocol. The scenarios include running a server behind a
NAT/Firewall, enabling direct connection between peers that exchange
addresses in an ad-hoc way, e.g. through an instant messaging
service, and enabling peer-to-peer communication with explicit
signaling, e.g. using SIP or H.323. These scenarios may include
several variants that we will present. We also present the evolution
of these scenarios when IPv6 provides global addresses, and
introduce the "6to4 router" scenario required for IPv6 transition,
and the IPSEC scenario enabled by IPv6.

Huitema                                                      [Page  1]

INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

The main purpose of this exercise is to explain the nature of the
holes that would have to be opened for the applications to work, so
as to derive the "functional requirements" of the firewall traversal
protocol. It is quite clear that there are other requirements,
notably security requirements. The scenarios described what the
application needs in order to run; whether the application is
actually allowed to run or not is a matter of local policy. In order
to meet the security requirements, the protocol will have to enable
adequate controls. In order to better understand how a security and
control can be applied, the scenarios include examples where
authentication is a gating operation.

This memo uses the definitions introduced in [MIDBOXFRAME], in
particular the definition of a Firewall/NAT.

2       Scenarios

In the following, we document a set of realistic scenarios that
should be enabled by a firewall traversal protocol. These scenarios
include:

*       Placing a server behind a firewall/NAT,

*       Enabling ad-hoc peer-to-peer applications,

*       Enabling peer-to-peer communication through explicit signaling
systems such as SIP or H.323.

We also take into account the deployment of IPv6, which introduces
variants of the previous scenarios, as well as new scenarios such as
the establishment of tunnels for carrying IPv6 packets through a
firewall, or the establishment of IPSEC associations between
internal and external hosts.

2.1     TCP server behind a firewall/NAT

An internal server wants to receive TCP-IP connections requests from
the outside (where outside is some place outside a domain). An
example is, running a web server in a domain protected by a
firewall.
   __________                           _________
  |          |-----[DNS Query]-------->|         |         ___________
  | external |<---[DNS Response]------<|   N.S.  |        |           |
  |   host   |                         |_________|        | Internal  |
  |__________|                                            |   Host    |
             v                                            |___________|
             v                          __________                v  ^
             v                         |          |<--midcom -----/  ^
             v                         | firewall |                  ^
             \>>>Connection Attempt>>>>|  / NAT   |>>>>>>>>>>>>>>>>>>^
                                       |__________|

Huitema                                                      [Page  2]

INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

In this scenario, the internal host publishes the IP address and TCP
port number at which it can be joined in a name server, using for
example SRV and A records in the DNS. The IP address that is
published must be valid in the "external" domain; if this external
domain is the current Internet, the published IP address must be
valid in the global Internet.

The scenario implies that the following operations happen in
sequence:

1)      The internal host interacts with the firewall/NAT, using the
midcom protocol. As a result of the interaction, the internal
host learns the IP address and TCP port that it may advertise to
external parties.

2)      The internal host publishes the information in a name server.

3)      The external host obtains the information from the name server.

4)      The external host issues a TCP connection request, and sends a
TCP SYN packet.

5)      The firewall/NAT receives the packet, performs address
translations and port mapping if necessary, and relays the TCP
SYN packet to the internal host.

6)      After that point, the TCP connection proceeds.

In the diagram, we depict only one external host, but this is an
example, not a limitation. Once the address and ports are published
in the name server, an unlimited number of external hosts may
attempt to connect to the internal server.

It is quite clear that the interaction between the internal host and
the firewall/NAT described in the first step of the scenario may
require some form of access control. The specific form of access
control will depend of the internal domain's policies. In practical
examples, these policies varies from allowing all internal hosts to
run services, in un-managed domains, to allowing only a specific set
of services on a specific set of hosts.

A particular result of the access control is that the first step of
this scenario may indeed fail, either because the opening of a hole
is not authorized or because the firewall/NAT does not have
sufficient resources. The internal host will explicitly learn that
it should not advertise any IP address or TCP port to third parties.
In this case, the host should not publish information, and external
hosts should not attempt to establish connections.

It is important that the behavior of the firewall/NAT be consistent:
if the mapping request at step 1 fails, then we expect that an

Huitema                                                      [Page  3]

INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

attempt to establish a connection from an external host will be
rejected; conversely, if the mapping request succeed, then we expect
the establishment of TCP connections to also succeed.

There are two interesting variants of that scenario: the use of UDP
instead of TCP, and the control of the firewall/NAT by a third party
instead of the internal host.

2.1.1    UDP Server behind a firewall/NAT

This scenario is exactly the same as the TCP server scenario, with
the difference that the external host issues unsolicited UDP
packets, instead of TCP/SYN packets. An example of this scenario is,
running a SIP server or a DNS server behind a NAT/Firewall.

2.1.2    TCP/UDP server authorized by a third party

This scenario differs from the base scenario in a simple way: the
midcom protocol is exercised by a third party instead of the host
itself. In the diagram, we call this third party a management
server:

   __________                           _________          ___________
  |          |-----[DNS Query]-------->|         |        |           |
  | external |<---[DNS Response]------<|   N.S.  |        | Management|
  |   host   |                         |_________|        |  Server   |
  |__________|                                            |___________|
             v                          __________                v
             v                         |          |<--midcom -----/
             v                         | firewall |        ___________
             \>>>Connection Attempt>>>>|  / NAT   |>>>>>>>| Internal  |
                                       |__________|       |   Host    |
                                                          |___________|

The management server will interact with the firewall/NAT, using the
midcom protocol, as in the step 1 of the main scenario. The other
steps will be unchanged. A key point of this scenario is that the
internal host is unaware of the midcom protocol; in practical
deployment, the internal host can be an unmodified server, such as a
web server responding to HTTP requests on incoming TCP connections,
or a DNS server responding to name requests on incoming UDP packets.

As in the original scenario, the mapping request may be rejected,
for example if the "management server" that attempts to establish
the mapping is not actually authorized to do so.

2.2     Peer-to-peer communication with ad-hoc rendezvous

The mediated peer-to-peer communication scenario describes hosts
that communicate through some external third party, such as an
instant messaging service, and then establish a direct communication
channel, such as a TCP connection. An example of this scenario is,

Huitema                                                      [Page  4]

INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

starting the exchange of files from an IM service.

   __________                    ___________
  |          |                  |           |
  | external |<-- Messaging --> | IM Server | <-- Messaging -.
  |   host   |                  |___________|              __v________
  |__________|                                            |           |
             v                                            |  internal |
             v                                            |   host    |
             v                                            |___________|
             v                          __________                v  ^
             v                         |          |<--midcom -----/  ^
             v                         | firewall |                  ^
             \>>>Connection Attempt>>>>|  / NAT   |>>>>>>>>>>>>>>>>>>^
                                       |__________|

This scenario does not involve any particular cooperation between
the firewall/NAT and the IM server. The connection between the
internal host and the IM system can use any protocol, in particular
combinations of TCP, HTTP and TLS.

The scenario implies that the following operations happen in
sequence:

1)      The internal and external hosts communicate through some form of
instant messaging service or chat room. At some point, they
decide to establish a direct channel, e.g. to exchange files.

2)      The internal host interacts with the firewall/NAT, using the
midcom protocol. As a result of the interaction, the internal
host learns the IP address and TCP port that it may advertise to
the external host.

3)      The internal host sends the IP address and the TCP port to the
external host.

4)      The external host issues a TCP connection request, and sends a
TCP SYN packet.

5)      The firewall/NAT receives the packet, performs address
translations and port mapping if necessary, and relays the TCP
SYN packet to the internal host.

6)      After that point, the TCP connection proceeds.

7)      After the application has finished using the connection, the
internal host may interact with the firewall/NAT and close the
hole.

In this scenario, the NAT firewall only has to authorize the
communication between a single internal host and a well identified
external host; the authorization typically only needs to remain

Huitema                                                      [Page  5]

INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

valid for a single TCP connection, or in any case for a limited
duration. The request in step 2 may be rejected by the firewall,
either for policy reasons, or because there are not sufficient
resource available; in this case, the peers should not attempt to
establish a connection. As we noted in scenario 2.1.1, it is
important that the behavior of the firewall/NAT be consistent: if
the mapping request at step 2 fails, then we expect that an attempt
to establish a connection at step 5 will be rejected; conversely, if
the mapping request succeeds, then we expect the establishment of
TCP connection to also succeed.

We expect that the decision to authorize the mapping request or not
will depend on a variety of parameters, such as the identity of the
internal user, the configuration of the internal system, the
identity of the external peer, the purpose of the connection, and
the amount of resource requested for the connection. The purpose of
the connection may be a generic notation such as "audio" or "video",
or a coded description of the application.

There are two variants of that scenario, when the dialog occurs over
UDP and when both hosts are hidden behind a firewall/NAT.


2.2.1    Peer-to-peer communication using UDP

This scenario is exactly the same as the TCP scenario, with the
difference that the external host issues UDP packets, instead of
TCP/SYN packets. An example of this scenario is, streaming audio or
video between two peers.

2.2.2    Both peers behind firewalls

When both peers are behind firewalls, it is hard to predict the IP
address that will be used by the host initiating the TCP connection.
In this situation, there are two options:

1)      Allow the internal host to accept TCP connections from any
external address.

2)      Let the "external" host use the midcom protocol to predict the
"external" IP address that it will use for the incoming
connection.

The first option may look insecure, but the possible insecurity of
accepting connections from multiple source is often mitigated by
application level protections, such as security tokens exchanged
through the IM channel. A variation of this option is to accept
connections from multiple sources, but restrict the hole to exactly
one source once the connection has been established. As in all other
scenarios, the firewall will have the option to accept or refuse the
requested hole; it is important that the confirmation or refusal be
explicit, and that the behavior of the firewall be consistent, i.e.

Huitema                                                      [Page  6]

INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

actually accept the connection if it accepted to open the hole.

2.3     Peer-to-peer communication with explicit signaling

In these scenarios, two peers that want to communicate use a
standard signaling protocol such as SIP or H.323. The communication
requests for internal host arrive to an internal server, e.g. the
"sip proxy" for the internal domain. In the diagram, we call this
agent the "internal server". The following description assumes the
use of SIP; scenarios that use an H.323 gatekeeper will use a
different message flow, but will involve similar interactions
between the gatekeeper and the firewall/NAT.

The scenarios imply that the internal server can receive signaling
packets from external hosts and servers. This is an application of
the previously described scenarios: TCP or UDP server behind a
firewall/NAT.

In these scenarios, the "internal server" has to understand the
location of the firewall/NAT in order to open the proper holes; this
may be difficult in big corporations with multiple firewalls, e.g.
in cases when the signaling flow will traverse a different
firewall/NAT than the media path. This will require some form of
firewall discovery; however, describing how discovery happens is
outside the scope of this document; the scenarios merely assume that
discovery somehow has happened, and that the server knows which
firewall/NAT will be used.

There are really two scenarios to consider, depending on whether the
call initiates from an internal host or from an external host. These
two scenarios assume that the firewall/NAT interacts with the
internal server. We will then consider a variant, in which the
interactions with the firewall/Nat are directly performed by the
internal host.

2.3.1    Explicit call from an internal host

In this scenario, an internal host calls a third party through the
internal server.


   __________                 _________             __________
  |          |<--[Invite]---<|         |<----------|          |<--.
  | external |---[response]->| Server  |---------->| Internal |--.|
  |   host   |               |_________|           |  Server  |  ||
  |__________|                                     |__________|  ||
            ^v                          __________           v   ||
            ^v                         |          |<--midcom-/   ||
            ^v                         | firewall |        ______v|___
            ^\>>> Media over UDP >>>>>>|  / NAT   |>>>>>>>| Internal  |
            \<<<<<<<<<<<<<<<<<<<<<<<<<<|__________|<<<<<<<|   Host    |
                                                          |___________|

Huitema                                                      [Page  7]

INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

The scenario implies that the following operations happen in
sequence:

1)      The internal host who want to start the call sends an invite
message to its preferred internal server. The invite message
carries the name of the invited user, and the IP address and UDP
ports through which the internal host intends to receive the
media, e.g. voice or video.

2)      The internal server determines that the target of the invite is
located outside the internal domain. If the firewall/NAT performs
address and port mapping, the internal server must interact with
the firewall/NAT and learn the "external mappings" corresponding
to the IP address and UDP ports used by the internal host.

3)      The internal server updates the address and port information in
the invite message, and relays the call to the "external server."

4)      The external server determines that the target of the invite is
located in a specific external host. It relays the call to this
host.

5)      The external host responds to the call. The response provides the
IP address and UDP port at which the external host will be
expecting to receive the media.

6)      The response message is relayed by the external server to the
internal server.

7)      The internal server receives the response. At this point, it
knows the IP addresses and ports used by both the internal and
the external host. The internal server interacts with the
firewall/NAT using the midcom protocol, to guarantee that the
exchange between the internal and external host will be
authorized.

8)      The response message is relayed by the internal server to the
internal host.

9)      The external and internal hosts send media packets to the
addresses and ports mentioned in the invite and response message;
these packets pass through the Firewall/NAT and reach their
destination.

We should note that, at step 2, the internal server must learn the
external mappings of the internal address and ports; at this stage,
it does no know the IP address and ports of the third party.

There is a potential race condition between the signaling message
that "responds to the call" and the first media packets sent by the
called party. Should the signaling loose the

Huitema                                                      [Page  8]

INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

race, the early media packets will bang against a closed firewall
and be clipped. It is in theory possible to design signaling
exchange that include a three ways handshake before media
transmission can start, followed by a message asking to start
ringing only after the availability of all necessary resource has
been verified. However, this is not compatible with existing
implementations of SIP or H.323, and would require a serious
revisiting of the gatewaying between SIP or H.323 and the telephone
network.

The description assumes that the hosts use the same UDP ports in
both direction of the media communication. This is not necessarily
the case. The source IP address may be unpredictable in the case of
multi-homed hosts; the source port may be systematically different
from the receive port in some implementation, e.g. parallel
processing of the send and receive channels by different software or
hardware components.

The scenario does not necessarily require a strict control by the
firewall/NAT of the source address and port authorized to send data.
Many implementations already support exchange of media level
authentication and encryption keys during the call set-up. This
provides a level of security that is at least as good as any control
of the source address and port: if attackers can manage to read the
signaling exchange and get the keys, they can just as well discover
the IP addresses and ports, and send forged packets.

As any other scenario, the firewall/NAT will have the option to
accept or refuse the requested hole. In this scenario, we observe
two successive interactions between the internal server and the
firewall/NAT: to request a mapping at step 2, to provide the address
of the external peer at step 7. It is important that the behavior of
the firewall/NAT be consistent, and that a hole opening authorized
at step 2 not be refused at step 5, when more details are available.
If the internal server learns early that the call will be refused,
it can terminate it without ever trying to "ring" the external peer.
If a call was first authorized and later refused, then the call will
proceed, the peer will be ringed and will accept the call, and only
at that point discover that there is no way to exchange media; this
is obviously very undesirable.

2.3.2    Explicit call to an internal host

In this scenario, a third party host calls an internal through the
internal server.








Huitema                                                      [Page  9]

INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

   __________                 _________             __________
  |          |---[Invite]--->|         |---------->|          |---.
  | external |<--[response]-<| Server  |<----------| Internal |<-.|
  |   host   |               |_________|           |  Server  |  ||
  |__________|                                     |__________|  ||
            ^v                          __________           v   ||
            ^v                         |          |<--midcom-/   ||
            ^v                         | firewall |        ______|v___
            ^\>>> Media over UDP >>>>>>|  / NAT   |>>>>>>>| Internal  |
            \<<<<<<<<<<<<<<<<<<<<<<<<<<|__________|<<<<<<<|   Host    |
                                                          |___________|


The scenario implies that the following operations happen in
sequence:

1)      The external host who want to start the call sends an invite
message to its preferred external server. The invite message
carries the name of the invited user, and the IP address and UDP
ports through which the external host intends to receive the
media, e.g. voice or video.

2)      The external server determines that the target of the invite is
located in the internal domain. It relays the call to the
"internal server."

3)      The internal server determines that the target of the invite is
located in a specific internal host. It relays the call to this
host.

4)      The internal host responds to the call. The response provides the
"internal" IP address and UDP port at which the internal host
will be expecting to receive the media, e.g. voice and video.

5)      The internal server receives the host's response. At this point,
it knows the IP addresses and ports used by both the internal and
the external host.

6)      The internal server interacts with the firewall/NAT using the
midcom protocol. If the firewall/NAT performs address mapping,
the internal server retrieves the mapping of the IP address(es)
and port(s) used by the internal host.

7)      The internal server prepares an updated response message that
reflects the mapping of the internal addresses. It sends the
response message to the external server.

8)      The response message is relayed to the external host by the
external server.

9)      The external and internal hosts send media packets to the
addresses and ports mentioned in the invite and response message;

Huitema                                                      [Page 10]


INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

these packets pass through the Firewall/NAT and reach their
destination.

We note that in this sequence the interaction with the firewall only
occurs after the internal host has accepted the call. This can
create an annoying effect if the interaction with the firewall
fails, equivalent to hearing a void telephone line after picking an
incoming call. To avoid this effect, the internal server will have
to somehow guarantee that the Firewall/NAT interaction will be
successful before relaying the call to the internal host. A possible
solution is to include the request of a provisional hole of some
sort at step 3, before the call is relayed to the internal host; if
the provisional hole is refused by the firewall/NAT, the internal
server can refuse the call without disturbing the internal user.

Just like scenario 2.3.1, it may not be possible or desirable to
predict or check the source IP address and UDP ports used by the
internal and external hosts.

2.3.3    Firewall interaction by the internal host

It is possible to update the previous two scenarios so that the
internal host interacts directly with the Firewall/NAT, rather than
relying on the internal server. This set-up has the advantage of
avoiding the "void telephone line" effect mentioned in the previous
scenario: the internal host that receives the invite can pick the
UDP ports used for audio and video and interact with the
firewall/NAT before "ringing" the user; if the interaction fails,
the call can be rejected without bothering the user. This set-up
however has the disadvantage that all internal hosts must become
able to interact with the Firewall/NAT, which in many cases may not
be practical.

The direct interaction between the internal host and the
NAT/Firewall is already described in the "peer-to-peer" scenarios of
the previous section. The only difference between these scenarios is
the possibility for the internal server to pass some form of
"authorization token" to the internal host.

2.3.4    Early media

The "early media" scenario is an important variations of the
scenario 2.3.1. Early media designates media transmission sent
before the actual completion of the call. Examples are ringing tones
and voice messages describing particular network conditions, such as
"we are trying to locate your correspondent." In the early media
scenario, the following interactions will happen in sequence:

1)      The internal host who want to start the call sends an invite
message to its preferred internal server, as in 2.3.1,

2)      The internal server determines that the target of the invite is

Huitema                                                      [Page 11]


INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

located outside the internal domain. If the firewall/NAT performs
address and port mapping, the internal server must interact with
the firewall/NAT and learn the "external mappings" corresponding
to the IP address and UDP ports used by the internal host, as in
2.3.1. In addition, the internal server requests the
authorization to receive packets from a yet unspecified external
source.

3)      The internal server updates the address and port information in
the invite message, and relays the call to the "external server."

4)      The external server, or a secondary server acting on its behalf,
sends a stream of voice packets towards the "external mappings"
of the IP address and UDP ports used by the internal host,

5)      The firewall/NAT receives these packets and forwards them to the
internal host,

6)      The call proceeds as in 2.3.1.

There is a common telephony practice of sending recorded
announcements during call set-up; the source IP address of these
announcements is not likely to be the same as the source IP address
used after call set-up is complete. It is theoretically possible
to use the equivalent of call transfer to switch between multiple
source in a controlled fashion, but this introduce a lot of
signaling complexity, and is incompatible with currently deployed
hardware and software. In practice, this scenario requires that the
firewall/NAT "opens a hole" without knowing the IP address and port
of the external peer.

2.3.5    Mobility of the external host

The mobility scenario can be thought as a complication of scenarios
2.3.1 or 2.3.2, in which the IP address of one of the peers is
allowed to change during a call, due to either mobility or network
renumbering. The scenario involves the following exchanges:

1)      The external host receives a new IP address, and sends a
signaling packet to the "internal server" mentioning the new IP
address,

2)      The internal server programs the firewall/NAT to start
authorizing packets between this new address and the internal
host,

3)      In parallel with 2, the internal server relays the signaling
message to the internal host,

4)      The internal and external hosts exchange packets with the new
address; the firewall/NAT authorizes these packets to proceed.


Huitema                                                      [Page 12]


INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

SIP supports that through the re-invite mechanism, but we should
note that there is either a gap in the call or a race condition
between media packets with the new source address and the signaling
message. The external host is likely to source packets with its new
address immediately after the address change; if the packets arrive
before the firewall/NAT has been programmed to accept them, the
packets will bang against the closed firewall/NAT and be dropped.

2.3.6    Multiple ports, port ranges

The SIP messages use the encoding defined in SDP [RFC2237] to
describe the IP addresses and TCP or UDP ports used my the various
media. In many cases, a single media stream will be spread over
multiple ports. SDP carries only one port number per media, and
states that "other ports used by the media application (such as the
RTCP port) should be derived algorithmically from the base media
port." When the media is transmitted using RTP [RFC1889], the choice
of the port number is very specific: "for UDP and similar protocols,
RTP uses an even port number and the corresponding RTCP stream uses
the next higher (odd) port number; if an application is supplied
with an odd number for use as the RTP port, it should replace this
number with the next lower (even) number." This obviously poses a
constraint to the allocation of ports and mappings by a NAT.

Most media streams are transmitted using a single pair of RTP and
RTCP ports. It is possible however to transmit a single media over
several RTP flows, for example using hierarchical encoding. In this
case, SDP will encode the port number used by RTP on the first flow,
and the number of flows, as in:

       m=video 49170/2 RTP/AVP 31

In this example, the media is sent over 2 consecutive pairs of
ports, corresponding respectively to RTP for the first flow (even
number, 49170), RTCP for the first flow (odd number, 49171), RTP for
the second flow (even number, 49172), and RTCP for the second flow
(odd number, 49173). This places a further constraint to any NAT
firewall traversal scheme: we must be able to ensure that a
consecutive range of N ports starting with an even number is mapped
to another consecutive range of N ports, also starting with an even
number.

2.4     IPv6 Scenarios

All of the scenarios mentioned above can be modified if the domains
have been upgraded to run IPv6. One difference between the IPv6 and
IPv4 scenarios is that the internal hosts can use global addresses;
however, there will also be cases in which address translation is
required after the introduction of IPv6, notably if one provides
interoperation between IPv6 and IPv4 using the NAT-PT scheme
[RFC2766]. When any form of address translation is required, e.g.
between IPv6 and IPv4 addresses, the scenarios are basically

Huitema                                                      [Page 13]


INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

unchanged; what may change is the content of the MIDCOM protocol
messages, which will have to include a mix of IPv4 and IPv6
addresses. On the other hand, when only global addresses are used in
the exchanges, the scenario are modified; the "middle box" does not
necessarily disappears, since in many domains there will still be
the need to perform explicit authorizations before letting data go
in and out; in these cases the "Firewall/NAT" combination becomes
strictly a "Firewall". In this section, we review how the
introduction of IPv6 and the use of global addresses can affect the
three classes of scenarios mentioned in the previous sections.

The transition to IPv6 will require the introduction of relay
routers, as specified in [RFC3056]; we discuss here how the MIDCOM
protocol can be used to open holes for the "tunnels" leading to
these relay routers.

In addition, the global addressing allows the introduction of
another scenario, the use of IPSEC between an internal and an
external host.

2.4.1    IPv6 TCP or UDP server behind a firewall

In this scenario, the internal host publishes the IP address and TCP
port number at which it can be joined in a name server, using for
example SRV and A6 records in the DNS. The sequence of operation is
the same as in the IPv4 case, but each of the step has a different
emphasis:

1)      The internal host interacts with the firewall, using the midcom
protocol. As a result of the interaction, the firewall learns the
IP address and TCP port that the host will use.

2)      The internal host publishes the information in a name server.

3)      The external host obtains the information from the name server.

4)      The external host issues a TCP connection request, and sends a
TCP SYN packet.

5)      The firewall receives the packet, checks that the destination
address and port are authorized, and relays the TCP SYN packet to
the internal host.

6)      After that point, the TCP connection proceeds.

The only reason for the first step in the scenario is access
control. If the domain's policy is to authorize all hosts to receive
all traffic, there is no need for this step - indeed, the firewall
becomes mostly a transparent IPv6 router. The impact of IPv6 on the
two variants of that scenario is obvious: the use of UDP will have
to be authorized if needed, and there may be a need to let a third
party perform the authorizations.

Huitema                                                      [Page 14]


INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

2.4.2    Peer-to-peer communication with ad-hoc rendezvous and IPv6

If both peers have a global IPv6 address, they will only have to
interact with a firewall if the domain's manager insists on having a
firewall control all incoming traffic; there will not be a need for
a NAT functionality. The internal host may still need to interact
with the firewall in order to "open a hole" for the packets coming
from the remote peer, but it will always be able to specify the
complete "five tuple" of protocol type, IP addresses and UDP ports;
the problem exposed in the case when both hosts were being firewalls
disappears.

2.4.3    Peer-to-peer communication with explicit signaling and IPv6

This scenario is also made simpler by the availability of global
addresses. In the case of a call from an internal host, the internal
server will not have to rewrite the addresses in the outgoing
"invite"; it will only have to interact with the firewall to open a
hole after the reception of the response. In the case of a call to
an internal host, the internal server may still have to interact
with a firewall if the domain managers insist on requiring this type
of protection; it will do so with an explicit knowledge of the IPv6
addresses and UDP ports used by both ends of the connection.

2.4.4    IPv6 transition service behind a firewall/NAT

A typical IPv6 transition scenario is described in [RFC3056]. In
this scenario, IPv6 is progressively made available by installing in
each site a "6to4" router, which receives IPv6 packets through
automatic tunnels and forwards them to internal IPv6 hosts.

   __________                      _________
  |          |-----[DNS Query]--->|         |         ___________
  | external |<---[DNS Response]-<|   N.S.  |        |           |
  |   host   |                    |_________|        |   6to4    |
  |__________|                                       |  Router   |
             ^                                       |___________|
             |                  __________            v   ^   ^
             |                 |          |<--midcom -/   |   |
             |                 | firewall |               |   \->
             \--- IPv6/IPv4 -->|  / NAT   |<-- IPv6/IPv4 -/   IPv6
                               |__________|


In this scenario, the 6to4 router provides the internal IPv6 hosts
with IPv6 addresses; the IPv6 prefix in these addresses is based on
a "global" IPv4 address of the domain. The IPv6 hosts will publish
their IPv6 addresses in the DNS. The external hosts will send IPv6
packets encapsulated in IPv4 headers, whose destination will be the
internal 6to4 router; the 6to4 router will receive the packets sent
by internal hosts to external hosts, and will encapsulate them with

Huitema                                                      [Page 15]


INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

adequate IPv4 headers.

The scenario implies that the following operations happen in
sequence:

1)      The 6to4 router interacts with the firewall/NAT, using the midcom
protocol. As a result of the interaction, the 6to4 router learns
a global IPv4 address that it can use to build a 6to4 prefix.

2)      The internal hosts publish IPv6 addresses based on this prefix in
a name server.

3)      The external host obtains the information from the name server.

4)      The external host sends IPv6 packets towards this address.

5)      The firewall/NAT receives the packet, notes that these are IPv6
packets carried in IPv4 (protocol type = 41), translates the
destination address if necessary and relays the packet to the
6to4 router.

6)      The 6to4 router removes the IPv4 header and forwards the IPv6
packet to the internal host.

7)      When the 6to4 router receives an IPv6 packet, it determines the
adequate IPv4 destination, and uses it to build an encapsulation
IPv4 header.

8)      The firewall/NAT receives the encapsulated packet. It may perform
translation of the source address if needed. It forwards the
packet to the IPv4 destination.

In the diagram, we depict only one external host, but this is an
example, not a limitation.

It is quite clear that, if fire walling function are desired for the
IPv6 traffic, these functions will have to be provided by the 6to4
router.

2.4.5    Enabling an IPSEC connection between IPv6 hosts

Once IPv6 provides global addresses to internal hosts, it becomes
possible to establish IPSEC associations between an internal host
and an external host. The establishment of the association will
start by a key exchange, and will continue with the exchange of
encrypted traffic.







Huitema                                                      [Page 16]


INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

   __________
  |          |
  | external |
  |   host   |                                             ___________
  |__________|                                            |           |
             ^                                            |  internal |
             |                                            |   host    |
             |                                            |___________|
             |                          __________                v  ^
             |                         |          |<--midcom -----/  |
             \-- Key exchange, IPSEC ->| firewall |<-----------------/
                                       |__________|

The scenario implies that the following operations happen in
sequence:

1)      The internal and external hosts decide to communicate, e.g. after
the internal host finds the address of the external host in the
DNS.

2)      The internal host and the external host exchange key negotiation
packets (IKE). The firewall passes these packets.

3)      The internal host uses the midcom protocol to signal to the
firewall that it is going to exchange encrypted traffic with an
external host, and obtains the authorization to proceed.

4)      IPSEC packets are exchanged.

5)      After the hosts have finished using the IPSEC association, the
internal host may interact with the firewall and close the hole.

We should note that this scenario requires that the firewall
delegates some of its control functions to the internal host:
encrypted traffic cannot be inspected.

As in all other scenarios, the firewall will have to explicitly
authorize the opening of a hole for the IPSEC association.

3       Security Considerations

Firewalls are used by domain managers to control the traffic that
can be exchanged between their domain and the Internet. In the
scenarios that we described, this control is relaxed in order to
enable certain applications. Relaxing the control has to be a
conscious decision of the domain manager.

4       IANA Considerations

The purpose of this memo is to document the allocation by IANA of an
IPv4 prefix dedicated to the 6to4 gateways to the native v6

Huitema                                                      [Page 17]


INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

Internet; there is no need for any recurring assignment.

5       Copyright

The following copyright notice is copied from RFC 2026 [Bradner,
1996], Section 10.4, and describes the applicable copyright for this
document.

Copyright (C) The Internet Society March 23, 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 assignees.

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.

6       Intellectual Property

The following notice is copied from RFC 2026 [Bradner, 1996],
Section 10.4, and describes the position of the IETF concerning
intellectual property claims made against this document.

The IETF takes no position regarding the validity or scope of any
intellectual property or other rights that might be claimed to
pertain to the implementation or use other technology described in
this document or the extent to which any license under such rights
might or might not be available; neither does it represent that it
has made any effort to identify any such rights.  Information on the
IETF's procedures with respect to rights in standards-track and
standards-related documentation can be found in BCP-11.  Copies of
claims of rights made available for publication and any assurances
of licenses to be made available, or the result of an attempt made
to obtain a general license or permission for the use of such

Huitema                                                      [Page 18]


INTERNET DRAFT               MIDCOM Scenarios             May 17, 2001

proprietary rights by implementers or users of this specification
can be obtained from the IETF Secretariat.

The IETF invites any interested party to bring to its attention any
copyrights, patents or patent applications, or other proprietary
rights which may cover technology that may be required to practice
this standard.  Please address the information to the IETF Executive
Director.

7       Acknowledgements

The discussion presented here was triggered by the meeting of the
MIDCOM working group in Minneapolis. An initial description of the
"TCP server" scenario was sent to the group's e-mail list by Eliot
Lear.

8       References

[RFC3056] B. Carpenter, K. Moore. "Connection of IPv6 Domains via
IPv4 Clouds." RFC 3056, February 2001.

[RFC2237] M. Handley, V. Jacobson, "SDP: Session Description
Protocol", RFC 2327, April 1998.

[RFC1889] H. Schulzrinne, S. Casner, R. Frederick, V. Jacobson. "RTP:
A Transport Protocol for Real-Time Applications", RFC 1889, January
1996.

[RFC2766] G. Tsirtsis, P. Srisuresh. "Network Address Translation -
Protocol Translation (NAT-PT)", RFC 2766, February 2000.

[MIDBOXFRAME] Middlebox Communication Architecture and Framework.
Work in progress.


9       Author's Address

Christian Huitema
Microsoft Corporation
One Microsoft Way
Redmond, WA 98052-6399

Email: huitema@microsoft.com










Huitema                                                      [Page 19]


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