Network Working Group M. Stenberg
Internet-Draft S. Paavolainen
Expires: January 12, 2001 T. Ylonen
T. Kivinen
SSH Communications Security Corp
July 14, 2000
IPsec NAT-Traversal
draft-stenberg-ipsec-nat-traversal-00.txt
Status of this Memo
This document is an Internet-Draft and is in full conformance with
all provisions of Section 10 of RFC2026.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that
other groups may also distribute working documents as
Internet-Drafts.
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.
This Internet-Draft will expire on January 12, 2001.
Copyright Notice
Copyright (C) The Internet Society (2000). All Rights Reserved.
Abstract
IPsec architecture is based on the concept of keeping data secure
while it is being transported across a network. Therefore there are
problems when packet headers change while in transmit across the
network, by virtue of NAT devices.
This draft details the changes needed in order to make both initial
IKE negotiation and subsequent authenticated/encrypted
communications across IPsec AH/ESP SAs work despite the changes in
the headers, and possible protocol transformations.
Stenberg, et. al. Expires January 12, 2001 [Page 1]
Internet-Draft IPsec NAT-Traversal July 2000
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 IPsec cases . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2.1 Host-to-host . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.2 Host-to-network . . . . . . . . . . . . . . . . . . . . . . 5
2.2.3 Network-to-network . . . . . . . . . . . . . . . . . . . . . 5
2.3 Issues stemming from NAT technology . . . . . . . . . . . . 5
2.4 Summary of issues . . . . . . . . . . . . . . . . . . . . . 5
3. Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.1 IKE probe . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.1.1 Determining of support . . . . . . . . . . . . . . . . . . . 7
3.1.2 NAT-Traversal need-probe . . . . . . . . . . . . . . . . . . 8
3.2 IPsec SA traffic encapsulation . . . . . . . . . . . . . . . 8
3.3 Heartbeat . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.3.1 Heartbeat format . . . . . . . . . . . . . . . . . . . . . . 11
3.4 Built-in NAT . . . . . . . . . . . . . . . . . . . . . . . . 11
4. Known issues with the solution . . . . . . . . . . . . . . . 13
4.1 Conceptual issues . . . . . . . . . . . . . . . . . . . . . 13
4.2 Overhead . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.3 Security considerations . . . . . . . . . . . . . . . . . . 14
4.4 Intellectual property rights . . . . . . . . . . . . . . . . 14
References . . . . . . . . . . . . . . . . . . . . . . . . . 15
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . 15
Full Copyright Statement . . . . . . . . . . . . . . . . . . 16
Stenberg, et. al. Expires January 12, 2001 [Page 2]
Internet-Draft IPsec NAT-Traversal July 2000
1. Introduction
NAT devices have proliferated recently. The eventual wider adoption
of IPv6 will also cause great deal of NAT activity, as IPv4 is here
to stay for the foreseeable future. Thus, there will be need for
bridging between IPv4 and IPv6 networks, and as long as there are
NATs around, basic IPsec as defined by RFC[1] will not work. It is
quite important that there is a defined standard for handling IPsec
traffic in networks with NAT devices. Preferably, a standard will
evolve to fit all possible cases that may arise.
In Section 2, most of the different IPsec+NAT permutations are
analyzed and finally, a list of issues is presented. Section 3
details the proposed solution to these issues. Finally, potential
problems in the solution are noted in Section 4.
1.1 Definitions
Following definitions will be used when discussing different types
of NATs.
o Static NAT: A NAT device that (for the traffic under discussion)
uses only a single static NAT translation.
o Dynamic NAT: A NAT device that (for the traffic under discussion)
creates address mappings dynamically based on some configured
static rules.
o Host NAT: A NAT that changes only the src/dst in packets.
o Host/port NAT: A NAT that changes src/dst and srcport/dstport in
packets.
o Protocol NAT: A NAT that changes the protocol of the packet; this
usually involves a whole new header for the packet.
Stenberg, et. al. Expires January 12, 2001 [Page 3]
Internet-Draft IPsec NAT-Traversal July 2000
2. Analysis
2.1 Assumptions
It can be safely assumed that IKE[2] works. IKE negotiations are
handled with normal UDP traffic, and therefore it should work
despite network address changes across the route. IP packet payloads
are assumed to be left unmodified; changes to the UDP headers can
occur, as long as nothing drops the packets before they reach the
host.
As normal IPsec traffic does not pass across host/port NATs (and may
not pass across protocol NATs), a complete NAT-Traversal design
should encapsulate IPsec SAs in UDP packets, which are (in most of
the important respects) like IP packets, except that they can pass
through all types of NATs.
2.2 IPsec cases
Initially, most of the different IPsec+NAT combinations are listed
here to make sure that all implications of NAT use are addressed.
IPsec cases can be divided to three different categories (with
possible NATs in various places along the route between hosts
employing IPsec).
o Host-to-host (tunnel or transport mode)
o Host-to-network (tunnel mode)
o Network-to-network (tunnel mode)
In all cases, the IKE responder must be, at best, only behind a
series of static host NATs; dynamic NATs do not work, for obvious
reasons (the IKE initiator cannot contact such an address), and
host/port NATs do not work because IKE is defined to be
port-500-only.
The IKE initiator can be behind any kind of NAT, although in cases
where initiation of traffic from both directions should be allowed
(primarily VPN-like cases), the same restrictions that apply to the
responder apply also to the initiator.
ISSUE0: Both hosts need to know that there is a NAT in the middle,
but currently IKE/IPsec do not provide such methodology (beyond the
fact that all IPsec SA packets, if they even arrive, will be dropped
as invalid).
ISSUE1: It is obvious that programs residing on an IKE responder
that is behind a host NAT cannot know about the existence of the
Stenberg, et. al. Expires January 12, 2001 [Page 4]
Internet-Draft IPsec NAT-Traversal July 2000
host NAT, nor about the specific address mappings configured there.
Thus, the IKE responder implementation should have advance knowledge
about the address mappings.
2.2.1 Host-to-host
Host-to-host traffic using tunnel or transport mode is the most
basic case; it only becomes interesting if there is no shared
address space between the parties (i.e., a VPN of sorts) and there
are NATs in between.
ISSUE2: If NATs are employed across the route, there may be
addressing conflicts in tunnel mode (and there WILL be conflicts in
transport mode). From the IKE responder point of view, the IKE
initiators' addresses may conflict if they are in private networks
(such as the IANA-assigned 10.0.0.0 subnet).
2.2.2 Host-to-network
Only tunnel mode is applicable for host-to-network communication,
and the only apparent problem is the potential lack of shared
address space (i.e., a host without an address in the remote network
that it is accessing). Therefore, there is potential for ISSUE2-type
of problems.
2.2.3 Network-to-network
Only tunnel mode is applicable in network-to-network communication,
and ISSUE2 is potentially a problem as well, although accounting for
network-to-network non-unique address mappings may be obscure.
2.3 Issues stemming from NAT technology
ISSUE3: The dynamic NATs may change their address mapping suddenly
(or they may be rebooted), making the remote host concept unworkable
even as a unique (host, port) pair.
2.4 Summary of issues
There are basically four problems that need addressing:
1. detection of network address translation during IKE negotiation
(ISSUE0),
2. a way of sending packets across the network so that NAT effects
can be countered, yet the security of the system will not be
affected (UDP encapsulation; assumption),
3. keeping NAT mapping static - NAT devices with dynamic host/port
Stenberg, et. al. Expires January 12, 2001 [Page 5]
Internet-Draft IPsec NAT-Traversal July 2000
allocation configurations typically contain timeouts that will
cause changes in addressing, if not circumvented by using a
heartbeat to keep the specific mappings up (ISSUE3), and
4. the lack of unique IP addresses in the NAT world; it is possible
for a server to have several clients with the same configured IP
address, although they appear to the server to be from different
hosts/ports (ISSUE2).
ISSUE1 (host NAT case, where the IKE responder does not know what
address to use) is trivial to solve, as seen in the end of Section
3.2.
Stenberg, et. al. Expires January 12, 2001 [Page 6]
Internet-Draft IPsec NAT-Traversal July 2000
3. Solution
The solution that resolves all the issues mentioned in Section 2.4
can be divided into four different parts, which are detailed in this
section:
o IKE probe to detect NAT presence,
o IPsec SA traffic encapsulation to counter NAT effects,
o NAT translation keepalive heartbeat which maintains NAT mappings,
and
o built-in NAT (if needed) to make addresses unique.
Incoming packet Outgoing packet
/ | | \
/ | | \
NAT-T decap. | | Un-NAT dst
| | | |
IPsec IPsec IPsec IPsec
| |
NAT src NAT-T encap.
Figure 1: IPsec processing with and without a NAT-Traversal process.
3.1 IKE probe
There is a need for two different exchanges in the IKE Phase 1
negotiation; initially, determining whether or not both sides
support NAT-Traversal. Then, if both sides do support it, there
should be a probe sequence that results in knowledge about whether
or not the network between hosts contains a NAT device.
As there is a need for two exchanges, of two messages each, it is
obvious that NAT-Traversal cannot be supported in P1 modes with less
than four messages sent across. Therefore, Aggressive Mode cannot be
used with this NAT-Traversal approach.
IKE Main Mode exchange contains 6 messages and therefore the probing
sequence can be done within it.
3.1.1 Determining of support
The NAT-Traversal capability of the remote host is determined by an
exchange of vendor strings; in Main Mode's four first messages, the
vendor id for this specification of NAT-Traversal
("draft-stenberg-ipsec-nat-traversal-00") MUST be sent if supported
Stenberg, et. al. Expires January 12, 2001 [Page 7]
Internet-Draft IPsec NAT-Traversal July 2000
(and it MUST be received by remote side) for the NAT-Traversal probe
to continue.
3.1.2 NAT-Traversal need-probe
Once the NAT-Traversal support of both parties has been determined,
in the last two encrypted messages of Main Mode, there are
additional private payloads sent in both directions.
Initially, in the 5th message of Main Mode, the initiator will add
one private payload to the message. PAYLOAD_TYPE (from private
range) is 211.
The payload should contain the following:
{perceived remote identity - IP address and port}
{one or more local identities - local interface address+port numbers}
Figure 2: Probe payload in Main Mode message #5
The probe payload is encoded as a series of Identification Payloads
of [3], with the perceived remote identity as the first payload, and
the local identities as the following payloads.
Once the IKE responder receives the payload represented in Figure 2,
the remote should check whether or not the remote identity, as
perceived by the IKE initiator, matches one of the locally
configured interface addresses (with proper port number). Also, the
remote identity as perceived by the IKE responder should match one
of the address+port pairs sent in the packet.
If one (or two) of those tests fails, the responder knows that
NAT-Traversal is needed. The decision about whether to use
NAT-Traversal or not is left up to the responder, and the responder
transmits the decision as a private payload of type 211 in the last
message of Main Mode.
The payload is just one byte long, and contains 0 when NAT-Traversal
is not elected to be used, and 1 when NAT-Traversal is chosen.
3.2 IPsec SA traffic encapsulation
Automatic use of NAT-Traversal encapsulation for IKE-negotiated
IPsec SAs MUST NOT be done. Instead, NAT-Traversal MUST be used only
when IKE negotiation has resulted in a decision to use
NAT-Traversal, or when manually keyed IPsec SAs are configured to
use it.
Traffic that is not in AH or ESP format MUST NOT be encapsulated
Stenberg, et. al. Expires January 12, 2001 [Page 8]
Internet-Draft IPsec NAT-Traversal July 2000
using this scheme, as that provides a way to create DDoS attacks,
and possibly some other security problems as well.
Normal AH/ESP traffic does not pass through NATs unmodified;
typically, the addresses may change (src/dst), and that makes the
resulting AH/ESP packet unusable. Thus, there has to be enough
redundant data to always be able to recreate a packet to its
original form. Additionally, it should preferably follow the same
NAT route as IKE packets, to make the implementation simpler.
Therefore (as noted before), the traffic has to be encapsulated as
UDP packets between two hosts (which implies that they follow same
route even in host/port NATs) using the IKE port. The basic idea
behind this NAT-Traversal data encapsulation format is that it
should be a format that can be adapted to future needs; therefore,
the only requirement for this initial version is that it contains a
version number, and it is invalid for IKE purposes.
An IPsec NAT-Traversal envelope for IPv4 packet encapsulation looks
like this:
<IP HEADER:
src=normal
dst=perceived remote host>
<UDP header:
srcport=500 (IKE port),
dstport=perceived remote port>
1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8
+---------------+---------------+---------------+---------------+
| NAT-T version | IP4 hlen | IP4 ToS | IP4 protocol |
+---------------+---------------+---------------+---------------+
| IP4 ID | IP4 frag_offset |
+---------------+---------------+---------------+---------------+
| IP4 src |
+---------------+---------------+---------------+---------------+
| <unused> |
+---------------+---------------+---------------+---------------+
| <unused> | 0 (IKE ver.) |
+---------------+---------------+
<IP4 HEADER leftovers, if any>
Figure 3: A NAT-Traversal envelope for an IPv4 IPsec packet.
The variables with the IP4 prefix are the original values that are
normally sent to the network as the header for the IPsec ESP data.
Stenberg, et. al. Expires January 12, 2001 [Page 9]
Internet-Draft IPsec NAT-Traversal July 2000
The header leftovers, if any, are the difference between the new
packet's IP4 header length and the original packet's IP4 header
length. The IP4 dst is not stored, as the remote host should be able
to know which address it is using to communicate with each host (in
the host-to-host case, with the responder behind a host NAT, the
only recourse is manual configuration data).
The NAT-T version field specifies the encapsulation specified;
defined types are as follows (and undefined types will be dropped
silently):
Name | Value
-------------------------------
TYPE_IPV4_ENCAPSULATION | 0x01
TYPE_HEARTBEAT | 0x02
The IPv4 encapsulation is defined in this section, and the heartbeat
is specified in Section 3.3.
Encapsulation occurs after IPsec processing, and it copies all
variables as-is from the original packet. Decapsulation occurs
before IPsec processing, and it copies values from the envelope to
the real packet and discards the envelope. In some cases, it may
involve changing of dst to be the host NAT address (if the remote
side negotiated with the host NAT address, not the real configured
address, and thus the host:spi pair is that of the host NAT
address:spi).
3.3 Heartbeat
Disclaimer: the IKE SA heartbeat should probably be used whenever
one becomes a standard. Until then, the NAT-Traversal will have its
own heartbeat that is entirely separate from the IKE SA and is used
between two hosts.
The sole purpose of the heartbeat is to keep the NATs in the network
route between hosts from removing the mapping from their dynamic
configuration (if any). Therefore, the actual contents of the
heartbeat can be more or less ignored (unless they stop arriving),
and thus encrypting them would serve no useful purpose.
Heartbeats MUST be sent as long as there is at least one IKE-probed
IPsec SA in existence between two hosts that employ NAT-Traversal to
communicate with each other.
The IKE initiator MUST send a heartbeat packet every
HEARTBEAT_INTERVAL (=10) seconds. The IKE responder SHOULD reply to
it. There MUST NOT be replies to address+port pairs with no IPsec
SAs up, or when there are too many heartbeat packets going on (i.e.,
Stenberg, et. al. Expires January 12, 2001 [Page 10]
Internet-Draft IPsec NAT-Traversal July 2000
there should be one reply [at most] in HEARTBEAT_INTERVAL/2
seconds). HEARTBEAT_INTERVAL MAY be higher, but as it is not
negotiated, both sides must be configured for a higher
HEARTBEAT_INTERVAL independently.
The heartbeat sequence in practice works as follows, using the
heartbeat format defined in Section 3.3.1:
Initiator Responder
-> Heartbeat (flag = FLAG_HEARTBEAT_PING)
<- Heartbeat (flag = FLAG_HEARTBEAT_REPLY)
3.3.1 Heartbeat format
The heartbeat packet format is very simple; it uses the same kind of
pseudo-IKE encapsulation as the previously defined IP4 envelope, but
with less fields in use.
1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8
+---------------+---------------+---------------+---------------+
| NAT-T version | HB flag | <unused> |
+---------------+---------------+---------------+---------------+
| <unused> |
+---------------+---------------+---------------+---------------+
| <unused> |
+---------------+---------------+---------------+---------------+
| <unused> |
+---------------+---------------+---------------+---------------+
| <unused> | 0 (IKE ver.) |
+---------------+---------------+
Figure 5: Heartbeat format.
Name | Value
----------------------------
FLAG_HEARTBEAT_PING | 0x01
FLAG_HEARTBEAT_REPLY | 0x02
Other values SHOULD be silently ignored.
3.4 Built-in NAT
Built-in host NAT implementation within the IPsec stack is necessary
in some tunnel cases and all transport cases. To stay consistent
with [2], which specifies that both tunnel and transport mode MUST
be supported, we define that there MUST be a built-in host NAT
implementation for NAT-Traversal use.
The built-in NAT is needed in some cases where ISSUE2 surfaces (see
Stenberg, et. al. Expires January 12, 2001 [Page 11]
Internet-Draft IPsec NAT-Traversal July 2000
Section 2.2 for details) to make the remote host(s) unique.
Typically, the host mapping should be from (perceived_remote_host,
perceived_remote_port) to some internal A- or B-class network.
Whenever the remote side successfully initiates IPsec SA employing
NAT-Traversal, there should be an internal NAT definition for the
(remote_host, remote_port) if one is required according to the local
configuration (or if transport mode is used, in which case internal
host NAT SHOULD always be employed). Whenever IPsec processing for
an incoming packet is done, the internal host NAT should be done to
the src. Whenever an outgoing packet headed towards an internal NAT
address enters the IPsec, the internal NAT address should be changed
to the address that was used for negotiating the IPsec SA.
In tunnel mode, it is possible that whole networks may need masking.
In the NAT-Traversal+IPsec case, a separate NAT box would not be
able to know about the (perceived_remote_host,
perceived_remote_port) pair which provides uniqueness to the
tunneled IP addresses. Therefore, there is a need for NAT within the
IPsec implementation. This MAY be supported, but no details about
implementation details will be provided here.
Stenberg, et. al. Expires January 12, 2001 [Page 12]
Internet-Draft IPsec NAT-Traversal July 2000
4. Known issues with the solution
4.1 Conceptual issues
Most of the solution is solid. An IPv6 and IPv4-IPv6
interoperability specification will be added to the next draft
version.
However, the non-unique hosts may cause problems, as there is a
potential problem of (host-port-proto-spi) not being unique any
more. The problem does not surface in the incoming traffic, but it
may occur in the outgoing case. There are (at least) a couple of
different solutions to the problem:
o Tying the remote-host,remote-port of NAT-T IPsec SA decapsulation
and the (host-port-proto-spi).
o Refusal of duplicate IPsec SA SPIs during IKE P2QM negotiation.
4.2 Overhead
Overall, this solution is almost the most minimal one possible that
covers most of the eventual possibilities and does not become overly
complex. Different types of overhead caused by this draft are noted
here, as well as possible ways of decreasing/removing the overheads
involved. Processing time and memory overhead are ignored as
negligible (some more processing for each packet, potentially
logarithmic searches for free addresses, minimal extra data for each
IPsec SA).
o IKE P1 negotiation extra payloads: Moderately small, typically
less than 200 bytes. Does not appear to be reducible.
o Each IPv4-based IPsec SA packet will contain extra overhead of 8
(UDP header) + 17 (NAT-T header) = 25 bytes. This might be
lowered slightly by dropping the IKE version field. Additionally,
more intelligent probing about what fields are changed across the
route would decrease the overhead as well, although at the cost
of increasing complexity. Some overhead is inevitable.
o Heartbeat overhead of 20 (IP header) + 8 (UDP header) + 17 (NAT-T
header) = 45 bytes * 2 every HEARTBEAT_INTERVAL. 9 bytes/second
may seem excessive, but as long as a general-purpose solution is
desired, it cannot be bypassed unless the encapsulation changes
from IKE-compliance (see previous entry). Heartbeat delay
negotiation for slower-reaction dynamic NAT routes (and disabling
when there are only static NATs) might be in order. This will be
investigated further in future versions of this draft.
Stenberg, et. al. Expires January 12, 2001 [Page 13]
Internet-Draft IPsec NAT-Traversal July 2000
4.3 Security considerations
Whenever changes to some fundamental parts of a security protocol
are proposed, the examination of security implications cannot be
skipped. Therefore, here are some observations regarding what is
affected, and whether or not the effect matters. This section will
be expanded further in future versions of this draft.
o IKE probe reveals NAT-Traversal support to everyone. This should
be a non-issue.
o IPsec encapsulation which contains source address before NAT is a
security leak of sorts, if internal network characteristics are
desired to be kept hidden.
o Obviously, the value of authentication mechanisms based on IP
addresses gets near zero once NATs are in the picture. That is
not necessarily a bad thing; for any real security, other
authentication measures than IP addresses should be used in any
case.
o Some DoS implications exist; a single malicious user can possibly
allocate up to (number-of-hosts-available) * 65535
(=number-of-ports-on-host) internal host IP addresses at the same
time - and cause that many negotiations as well (this is 65535
times as much DoS potential as traditional IKE). As the IP
addresses are allocated only after authentication is successful,
the culprit is known, however, and therefore this can be
considered a slight risk at best.
o The encapsulation scheme prevents some control of IP-level
headers. Therefore, there is some potential for forging of some
fields of the ESP transport mode IP header. Because the
decapsulated destination+spi MUST remain unchanged, there are no
apparent security risks, unless the remote end's IP-level
handling has some exploitable bugs (and even then only in
implementation approach B of Appendix A of [1]). AH+ESP should be
employed if IP-level header integrity is desired, as usual.
Therefore, this is a non-issue; although firewall administration
loses some control over IP headers that are passed through, use
of flawed IP protocol implementations is in itself a bad idea.
4.4 Intellectual property rights
SSH Communications Security Corp has patent applications which may
cover parts of this technology. If this technology starts to
progress on the IETF standards track, SSH is willing to seek a
licensing solution that allows widespread use of this technology.
Stenberg, et. al. Expires January 12, 2001 [Page 14]
Internet-Draft IPsec NAT-Traversal July 2000
References
[1] Kent, S. and R. Atkinson, "Security Architecture for the
Internet Protocol", RFC 2401, November 1998.
[2] Harkins, D. and D. Carrel, "The Internet Key Exchange (IKE)",
RFC 2409, November 1998.
[3] Piper, D., "The Internet IP Security Domain of Interpretation
for ISAKMP", RFC 2407, November 1998.
Authors' Addresses
Markus Stenberg
SSH Communications Security Corp
Fredrikinkatu 42
FIN-00100 Helsinki
Finland
EMail: mstenber@ssh.com
Santeri Paavolainen
SSH Communications Security Corp
Fredrikinkatu 42
FIN-00100 Helsinki
Finland
EMail: santtu@ssh.com
Tatu Ylonen
SSH Communications Security Corp
Fredrikinkatu 42
FIN-00100 Helsinki
Finland
EMail: ylo@ssh.com
Tero Kivinen
SSH Communications Security Corp
Fredrikinkatu 42
FIN-00100 Helsinki
Finland
EMail: kivinen@ssh.com
Stenberg, et. al. Expires January 12, 2001 [Page 15]
Internet-Draft IPsec NAT-Traversal July 2000
Full Copyright Statement
Copyright (C) The Internet Society (2000). 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.
Stenberg, et. al. Expires January 12, 2001 [Page 16]