Internet Engineering Task Force               Robert E. Gilligan
INTERNET-DRAFT                                Erik Nordmark
                                              Sun Microsystems, Inc.

                                              March 17,

                                              May 15, 1995

           Transition Mechanisms for IPv6 Hosts and Routers
                 <draft-ietf-ngtrans-trans-mech-00.txt>
                 <draft-ietf-ngtrans-trans-mech-01.txt>

Abstract

This document specifies IPv4 compatibility mechanisms that can be
implemented by IPv6 hosts and routers.  These mechanisms include
providing complete implementations of both versions of the Internet
Protocol (IPv4 and IPv6), and tunneling IPv6 packets over IPv4 routing
infrastructures.  They are designed to allow IPv6 nodes to maintain
complete compatibility with IPv4, which should greatly simplify the
deployment of IPv6 in the Internet, and facilitate the eventual
transition of the entire Internet to IPv6.

Status of this Memo

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.''

To learn the current status of any Internet-Draft, please check the
``1id-abstracts.txt'' listing contained in the Internet- Drafts Shadow
Directories on ds.internic.net (US East Coast), nic.nordu.net (Europe),
ftp.isi.edu (US West Coast), or munnari.oz.au (Pacific Rim).

This Internet Draft expires on September 17, November 15, 1995.

1. Introduction

This specification defines mechanisms that IPv6 hosts and routers may
implement

The key to be compatible a successful IPv6 transition is compatibility with the large
installed base of IPv4 hosts and routers.  Maintaining compatibility
with IPv4 while deploying IPv6 will streamline the task of transitioning
the Internet to IPv6.  This specification defines a set of mechanisms
that IPv6 hosts and routers may implement in order to be compatible with
IPv4 hosts and routers.

The mechanisms in this document are designed to be employed by IPv6
hosts and routers that need to interoperate with IPv4 hosts and utilize
IPv4 routing infrastructures.  We expect that complete compatibility
with IPv4 will be necessary most nodes in the Internet
will need such compatibility for a long time to come, and perhaps even
indefinitely.

However, IPv6 may be used in some environments where interoperability
with IPv4 is not required.  IPv6 nodes that are designed to be used in
such environments need not use or even implement these mechanisms.

The mechanisms specified here include:

   -    Dual IP layer.  Providing complete support for both IPv4 and
        IPv6 in hosts and routers.

   -    IPv6 over IPv4 tunneling.  Encapsulating IPv6 packets within
        IPv4 headers to carry them over IPv4 routing infrastructures.
        Two types of tunneling are employed: configured and automatic.

Additional transition and compatibility mechanisms may be developed in
the future.  These will be specified in other documents.

1.2. Terminology

The following terms are used in this document:

Types of Nodes

        IPv4-only node:

                A  host  or  router  that  implements  only  IPv4.    An
                IPv4-only  node does not understand IPv6.  The installed
                base of IPv4  hosts  and  routers  existing  before  the
                transition begins are IPv4-only nodes.

        IPv6/IPv4 node:

                A host or router that implements both IPv4 and IPv6.

        IPv6-only node:

                A host or router that implements IPv6, and does not
                implement IPv4.  The operation of IPv6-only nodes is not
                addressed here.

        IPv6 node:

                Any host or router that implements IPv6.  IPv6/IPv4 and
                IPv6-only nodes are both IPv6 nodes.

        IPv4 node:

                Any host or router that implements IPv4.  IPv6/IPv4 and
                IPv4-only nodes are both IPv4 nodes.

Types of IPv6 Addresses

        IPv4-compatible IPv6 address:

                An IPv6 address, assigned to an IPv6/IPv4 node, which
                bears the high-order 96-bit prefix 0:0:0:0:0:0, and an
                IPv4 address in the low-order 32-bits.  IPv4-compatible
                addresses are used by the automatic tunneling mechanism.

        IPv6-only address:

                The remainder of the IPv6 address space.  An IPv6
                address that bears a prefix other than 0:0:0:0:0:0.

Techniques Used in the Transition

        IPv6-over-IPv4 tunneling:

                The technique of encapsulating IPv6 packets within IPv4
                so that they can be carried across IPv4 routing
                infrastructures.

        IPv6-in-IPv4 encapsulation:

                IPv6-over-IPv4 tunneling.

        Configured tunneling:

                IPv6-over-IPv4 tunneling where the IPv4 tunnel endpoint
                address is determined by configuration information on
                the encapsulating node.

        Automatic tunneling:

                IPv6-over-IPv4 tunneling where the IPv4 tunnel endpoint
                address is determined from the IPv4 address embedded in
                the IPv4-compatible destination address of the IPv6
                packet.

1.3. Structure of this Document

The remainder of this document is organized into three sections:

   -    Section 2 discusses the IPv4-compatible address format.

   -    Section 3 discusses the operation of nodes with a dual IP
        layer, IPv6/IPv4 nodes.

   -    Section 4 discusses IPv6-over-IPv4 tunneling.

2. Addressing

The automatic tunneling mechanism uses a special type of IPv6 address,
termed an "IPv4-compatible" address.  An IPv4-compatible address is
identified by an all-zeros 96-bit prefix, and holds an IPv4 address in
the low-order 32-bits.  IPv4-compatible addresses are structured as
follows:

        |              96-bits                 |   32-bits    |
        +--------------------------------------+--------------+
        |            0:0:0:0:0:0               | IPv4 Address |
        +--------------------------------------+--------------+

                 IPv4-Compatible IPv6 Address Format

IPv4-compatible addresses are assigned to IPv6/IPv4 nodes that support
automatic tunneling.  Nodes that are configured with IPv4-compatible
addresses may use the complete address as their IPv6 address, and use
the embedded IPv4 address as their IPv4 address.

The remainder of the IPv6 address space (that is, all addresses with
96-bit prefixes other than 0:0:0:0:0:0) are termed "IPv6-only
Addresses."

3. Dual IP Layer

The most straightforward way for IPv6 nodes to remain compatible with
IPv4-only nodes is by providing a complete IPv4 implementation.  IPv6
nodes that provide a complete IPv4 implementation in addition to their
IPv6 implementation are called "IPv6/IPv4 nodes."  IPv6/IPv4 nodes have
the ability to send and receive both IPv4 and IPv6 packets.  They can
directly interoperate with IPv4 nodes using IPv4 packets, and also
directly interoperate with IPv6 nodes using IPv6 packets.

The dual IP layer technique may or may not be used in conjunction with
the IPv6-over-IPv4 tunneling techniques, which are described in
section 4.  An IPv6/IPv4 node that supports tunneling may support only
configured tunneling, or both configured and automatic tunneling.
Thus three configurations are possible:

   -    IPv6/IPv4 node that does not perform tunneling.

   -    IPv6/IPv4 node that performs configured tunneling only.

   -    IPv6/IPv4 node that performs configured tunneling and
        automatic tunneling.

3.1. Address Configuration

Because they support both protocols, IPv6/IPv4 nodes may be configured
with both IPv4 and IPv6 addresses.  Although the two addresses may be
related to each other, this is not required.  IPv6/IPv4 nodes may be
configured with IPv6 and IPv4 addresses that are unrelated to each
other.

Nodes that perform automatic tunneling are configured with
IPv4-compatible IPv6 addresses.  These may be viewed as single
addresses that can serve both as IPv6 and IPv4 addresses.  The entire
128-bit IPv4-compatible IPv6 address is used as the node's IPv6
address, while the IPv4 address embedded in low-order 32-bits serves
as the node's IPv4 address.

IPv6/IPv4 nodes may use the stateless IPv6 address configuration
mechanism [5] or DHCP for IPv6 [3] to acquire their IPv6 address.
These mechanisms may provide either IPv4-compatible or IPv6-only IPv6
addresses.

IPv6/IPv4 nodes may use IPv4 mechanisms to acquire their IPv4
addresses.

IPv6/IPv4 nodes that perform automatic tunneling may also acquire their
IPv4-compatible IPv6 addresses from another source: IPv4 address

configuration protocols.  A node may use any IPv4 address configuration
mechanism to acquire its IPv4 address, then "map" that address into an
IPv4-compatible IPv6 address by pre-pending it with the 96-bit prefix
0:0:0:0:0:0.  This mode of configuration allows IPv6/IPv4 nodes to
"leverage" the installed base of IPv4 address configuration servers.  It
can be particularly useful in environments where IPv6 routers and
address configuration servers have not yet been deployed.

The specific algorithm for acquiring an IPv4-compatible address using
IPv4-based address configuration protocols is as follows:

   1)   The IPv6/IPv4 node uses standard IPv4 mechanisms or protocols
        to acquire its own IPv4 address.  These include:

           -    The Dynamic Host Configuration Protocol (DHCP) [2]
           -    The Bootstrap Protocol (BOOTP) [1]
           -    The Reverse Address Resolution Protocol (RARP) [9]
           -    Manual configuration
           -    Any other mechanism which accurately yields the node's
                own IPv4 address

   2)   The node uses this address as its IPv4 address.

   3)   The node prepends the 96-bit prefix 0:0:0:0:0:0 to the 32-bit
        IPv4 address that it acquired in step (1).  The result is an
        IPv4-compatible IPv6 address with the node's own IPv4-address
        embedded in the low-order 32-bits.  The node uses this address
        as its own IPv6 address.

3.1.1. IPv4 Loopback Address

Many IPv4 implementations treat the address 127.0.0.1 as a "loopback
address" -- an address to reach services located on the local machine.
Per the host requirements specification [11], section 3.2.1.3, IPv4
packets addressed from or to the loopback address are not to be sent
onto the network; they must remain entirely within the node.  IPv6/IPv4
implementations may treat the IPv4-compatible IPv6 address ::127.0.0.1
as an IPv6 loopback address.  Packets with this address should also
remain entirely within the node, and not be transmitted onto the
network.

3.2.  DNS

The Domain Naming System (DNS) is used in both IPv4 and IPv6 to map
hostnames into addresses.  A new resource record type named "AAAA" has
been defined for IPv6 addresses [6].  Since IPv6/IPv4 nodes must be able
to interoperate directly with both IPv4 and IPv6 nodes, they must must
provide resolver libraries capable of dealing with IPv4 "A" records as

well as IPv6 "AAAA" records.

Some sites use local host tables instead of, or in addition to, the
DNS.  Use of host tables may be particularly useful in the very early
stages of transition before the DNS infrastructure has been converted
to support AAAA records.  Therefore, implementations may provide a
host table mechanism in addition to their DNS resolver.

Note that the local host table mechanism does not scale very well, so
its use is not recommended for large sites.  Further discussion of the
host table issue can be found in section 6.1.1 of "Requirements for
Internet Hosts -- Application and Support" [10].

3.2.1.  Handling Records for IPv4-Compatible Addresses

When an IPv4-compatible IPv6 addresses is assigned to an IPv6/IPv4
host that supports automatic tunneling, both A and AAAA records are
listed in the DNS.  The AAAA record holds the full IPv4-compatible
IPv6 address, while the A record holds the low-order 32-bits of that
address.  The AAAA record is needed so that queries by IPv6 hosts can
be satisfied.  The A record is needed so that queries by IPv4-only
hosts, whose resolver libraries only support the A record type, will
locate the host.

DNS resolver libraries on IPv6/IPv4 nodes must be capable of handling
both AAAA and A records.  However, when a query locates an AAAA record
holding an IPv4-compatible IPv6 address, and an A record holding the
corresponding IPv4 address, the resolver library need not necessarily
return both addresses.  It has three options:

   -    Return only the IPv6 address to the application.

   -    Return only the IPv4 address to the application.

   -    Return both addresses to the application.

The selection of which address type to return in this case, or, if
both addresses are returned, in which order they are listed, can
affect what type of IP traffic is generated.  If the IPv6 address is
returned, the node will communicate with that destination using IPv6
packets (in most cases encapsulated in IPv4); If the IPv4 address is
returned, the communication will use IPv4 packets.

The way that DNS resolver implementations handle redundant records for
IPv4-compatible addresses may depend on whether that implementation
supports automatic tunneling, or whether it is enabled.  For example, an
implementation that does not support automatic tunneling would not
return IPv4-compatible IPv6 addresses to applications because those

destinations are generally only reachable via tunneling.  On the other
hand, those implementations in which automatic tunneling is supported
and enabled may elect to return only the IPv4-compatible IPv6 address
and not the IPv4 address.

4. IPv6-over-IPv4 Tunneling

In most deployment scenarios, the IPv6 routing infrastructure will be
built up over time.  While the IPv6 infrastructure is being deployed,
the existing IPv4 routing infrastructure can remain functional, and
can be used to carry IPv6 traffic.  Tunneling provides a way to
utilize an existing IPv4 routing infrastructure to carry IPv6 traffic.

IPv6/IPv4 hosts and routers can tunnel IPv6 datagrams over regions of
IPv4 routing topology by encapsulating them within IPv4 packets.
Tunneling can be used in a variety of ways:

   -    Router-to-Router.  IPv6/IPv4 routers interconnected by an IPv4
        infrastructure can tunnel IPv6 packets between themselves.  In
        this case, the tunnel spans one segment of the end-to-end path
        that the IPv6 packet takes.

   -    Host-to-Router.  IPv6/IPv4 hosts can tunnel IPv6 packets to an
        intermediary IPv6/IPv4 router that is reachable via an IPv4
        infrastructure.  This type of tunnel spans the first segment
        of the packet's end-to-end path.

   -    Host-to-Host.  IPv6/IPv4 hosts that are interconnected by an
        IPv4 infrastructure can tunnel IPv6 packets between
        themselves.  In this case, the tunnel spans the entire
        end-to-end path that the packet takes.

   -    Router-to-Host. IPv6/IPv4 routers can tunnel IPv6 packets to
        their final destination IPv6/IPv4 host.  This tunnel spans
        only the last segment of the end-to-end path.

Tunneling techniques are usually classified according to the mechanism
by which the encapsulating node determines the address of the node at
the end of the tunnel.  In the first two tunneling methods listed above
-- router-to-router and host-to-router -- the IPv6 packet is being
tunneled to a router.  The endpoint of this type of tunnel is an
intermediary router which must decapsulate the IPv6 packet and forward
it on to its final destination.  When tunneling to a router, the
endpoint of the tunnel is different from the destination of the packet
being tunneled.  So the addresses in the IPv6 packet being tunneled do
not provide the IPv4 address of the tunnel endpoint.  Instead, the
tunnel endpoint address must be determined from configuration
information on the node performing the tunneling.  We use the term
"configured tunneling" to describe the type of tunneling where the
endpoint is explicitly configured.

In the last two tunneling methods -- host-to-host and router-to-host
-- the IPv6 packet is tunneled all the way to its final destination.

The tunnel endpoint is the node to which the IPv6 packet is addressed.
Since the endpoint of the tunnel is the destination of the IPv6
packet, the tunnel endpoint can be determined from the destination
IPv6 address of that packet: If that address is an IPv4-compatible
address, then the low-order 32-bits hold the IPv4 address of the
destination node, and that can be used as the tunnel endpoint address.
This technique avoids the need to explicitly configure the tunnel
endpoint address.  Deriving the tunnel endpoint address from the
embedded IPv4 address of the packet's IPv6 address is termed
"automatic tunneling".

The two tunneling techniques -- automatic and configured -- differ
primarily in how they determine the tunnel endpoint address.  Most of
the underlying mechanisms are the same:

   -    The entry node of the tunnel (the encapsulating node) creates an
        encapsulating IPv4 header and transmits the encapsulated packet.

   -    The exit node of the tunnel (the decapsulating node) receives
        the encapsulated packet, removes the IPv4 header, updates the
        IPv6 header, and processes the received IPv6 packet.

   -    The encapsulating node may need to maintain soft state
        information for each tunnel recording such parameters as the MTU
        of the tunnel its path length in order to correctly generate process IPv6 ICMP error messages. packets forwarded into
        the tunnel.  Since the number of tunnels that any one host or
        router may be using may grow to be quite large, this state
        information can be cached and discarded when not in use.

The next section discusses the common mechanisms that apply to both
types of tunneling.  Subsequent sections discuss how the tunnel endpoint
address is determined for automatic and configured tunneling.

4.1. Common Tunneling Mechanisms

The encapsulation of an IPv6 datagram in IPv4 is shown below:

                                        +-------------+
                                        |    IPv4     |
                                        |   Header    |
        +-------------+                 +-------------+
        |    IPv6     |                 |    IPv6     |
        |   Header    |                 |   Header    |
        +-------------+                 +-------------+
        |  Transport  |                 |  Transport  |
        |   Layer     |      ===>       |   Layer     |
        |   Header    |                 |   Header    |
        +-------------+                 +-------------+
        |             |                 |             |
        ~    Data     ~                 ~    Data     ~
        |             |                 |             |
        +-------------+                 +-------------+

                      Encapsulating IPv6 in IPv4

In addition to adding an IPv4 header header, the encapsulating node also has to
handle some more complex issues:

  -     Determine when to fragment and when to report an ICMP "packet
        too big" error back to the source.

  -     How to account for the tunnel in the IPv6 Hop Limit field.

  -     How to reflect IPv4 ICMP errors from routers along the tunnel
        path back to the source as IPv6 ICMP errors.

Those issues are discussed in the following sections.

4.1.1.  Tunnel MTU and fragmentation Fragmentation

The encapsulating node could view encapsulation as IPv6 using IPv4 as a
link layer with a very large MTU (65535-20 bytes to be exact; 20 bytes
"extra" are needed for the encapsulating IPv4 header).  The
encapsulating node would need only to report IPv6 ICMP "packet too big"
errors back to the source for packets that exceed this MTU.  However,
such a scheme would be inefficient for two reasons:

  1)    It would result in more fragmentation than needed. IPv4 layer
        fragmentation should be avoided due to the performance problems
        caused by the loss unit being smaller than the retransmission
        unit.
        unit [13].

  2)    Any IPv4 fragmentation occurring inside the tunnel would have to
        be reassembled at the tunnel endpoint.  For tunnels that
        terminate at a router, this would require additional memory to
        reassemble the IPv4 fragments into a complete IPv6 packet before
        that packet could be forwarded onward.

The fragmentation inside the tunnel can be reduced to a minimum by
having the encapsulating node track the IPv4 Path MTU across the tunnel
(using tunnel,
using the IPv4 Path MTU Discovery Protocol [8] and recording the
resulting path MTU
in the internet layer). MTU.  The IPv6 layer in the encapsulating node can then
view a tunnel as a link layer with an MTU equal to the IPv4 path MTU,
minus the size of the encapsulating IPv4 header.

Note that this does not completely eliminate IPv4 fragmentation in the
case when the IPv4 path MTU would result in an IPv6 MTU less than 576
bytes. (Any link layer used by IPv6 has to have an MTU of at least 576
bytes [4].) In this case the IPv6 layer has to "see" a link layer
with an MTU of 576 bytes and the encapsulating node has to use IPv4
fragmentation in order to forward the 576 byte IPv6 packets.

The encapsulating node can employ the following algorithm to determine
when to forward an IPv6 packet that is larger than the tunnel's path MTU
using IPv4 fragmentation, and when to return an IPv6 ICMP "packet too
big" message:

        if (IPv4 path MTU - 20) is less than or equal to 576
                if packet is larger than 576 bytes
                        Send IPv6 ICMP "packet too big" with MTU = 576.
                        Drop packet.
                else
                        Encapsulate but do not set the Don't Fragment
                        flag in the IPv4 header. The resulting IPv4
                        packet might be fragmented by the IPv4 layer on
                        the encapsulating node. node or by some router along
                        the IPv4 path.
                endif
        else
                if packet is larger than (IPv4 path MTU - 20)
                        Send IPv6 ICMP "packet too big" with
                        MTU = (IPv4 path MTU - 20).
                        Drop packet.
                else
                        Encapsulate and set the Don't Fragment flag
                        in the IPv4 header.
                endif
        endif

Encapsulating nodes that have a large number of tunnels might not be
able to store the IPv4 Path MTU for all tunnels. Such nodes can, at the
expense of additional fragmentation in the network, avoid using the IPv4
Path MTU algorithm across the tunnel and instead use the MTU of the link
layer (under IPv4) in the above algorithm instead of the IPv4 path MTU.

In this case the Don't Fragment bit must not be set in the encapsulating
IPv4 header.

4.1.2.  Hop Limit

The IPv4 hops of an

IPv6-over-IPv4 tunnel can be accounted for in one of
two ways:

   1)   Each of the "hops" that an encapsulated IPv6 datagram takes
        through IPv4 routers can be reflected in tunnels are modeled as "single-hop".  That is, the IPv6
hop limit
        field.  For example, if the IPv4 path length of a tunnel is 5
        hops, the IPv6 "hop limit" field is decremented by 5 1 when an IPv6 packet travels through traverses the tunnel.  We use the term
        "multi-hop" to describe tunnels that use this model.

   2)   The tunnel can be modeled as consuming only one IPv6 hop
        independent of its IPv4 path length.  That is, the IPv6 hop
        limit is decremented only by 1 when an IPv6 packet traverses the
        tunnel.  We use the term "single-hop" to describe tunnels that
        use this model.

These two models can be used to achieve different objectives.  The
multi-hop model can be useful to enforce the scope limitations imposed
by the sender of the IPv6 datagram.  It also makes the tunnel
"traceroute detectable": by sending IPv6 packets with hop limit values
that will cause them to "expire" within the tunnel, network management
programs like "traceroute" can locate tunnels and determine their path
length.  Such programs can not determine the addresses of the IPv4
routers within the tunnel, however.
The single-hop model is useful if the administrator wishes serves to hide the existence of a tunnel.  Since a single-hop  The
tunnel only "consumes" one
IPv6 hop, it is opaque to users of the network, and is not detectable by programs like
network diagnostic tools such as traceroute.

The multi-hop single-hop model can be is implemented by having the encapsulating node
copy and
decapsulating nodes process the IPv6 hop limit into the IPv4 TTL field when it composes the
encapsulating packet, and having the decapsulating node copy the IPv4
TTL field back into as they would if
they were forwarding a packet on to any other datalink.  That is, they
decrement the IPv6 hop limit field.

The single-hop model is implemented by having the encapsulating node
select the IPv4 TTL independently of the 1 when forwarding an IPv6 hop limit, and the
decapsulating packet. (The
originating node and final destination do not copying decrement the IPv4 hop limit.)

The TTL into of the encapsulating IPv4 header is selected in an
implementation dependent manner.  The current suggested value is
published in the hop limit
field. "Assigned Numbers RFC.  Implementations may provide either model or both.  Implementations that
provide both models may wish a
mechanism to give administrators allow the ability administrator to configure which model is used for each tunnel.

If implementations provide configurability, it is important that both
ends of the tunnel -- the encapsulating and decapsulating nodes -- are
configured to use the same model.  If the tunnel endpoints are

configured differently, packets could end up with an incorrect IPv6 hop
limit.

No serious problems would result if the encapsulating node were
configured to use the multi-hop model, but the decapsulating node was
configured to use the single-hop model.  The results would the same as
if both ends were configured to use the single-hop model.  However, two
failure modes can occur if the encapsulating node is configured to use
the single-hop model and the decapsulating node is configured to use the
multi-hop model:

   -    The IPv6 packet exits the tunnel with a larger hop limit than it
        had when entering the tunnel.  This would occur if the amount of
        IPv4 TTL remaining when the packet reached the decapsulating
        node was larger than the IPv6 hop count.  This failure can be
        thought of as the IPv6 packet "gaining hop limit" when passing
        through the tunnel.

   -    The number of IPv6 hops "consumed" in passing through the tunnel
        is more than IPv4 path length of the tunnel.  This would occur
        if the difference between the IPv6 hop limit in the packet and
        the remaining IPv4 TTL was greater than the IPv4 path length of
        the tunnel.  This failure can be thought of as the IPv6 packet
        "loosing too much hop limit" when passing through the tunnel.

Note that in both of these cases, the original IPv6 hop limit is lost.
Its value after transiting the tunnel is related only to the IPv4 TTL
selected by the encapsulating node, which is not related to the hop
limit in the IPv6 packet.

Of the two potential failure modes above, the first is more serious
since it could cause a packet to "live forever".  A routing loop which
sent IPv6 packets through such a tunnel could cause an infinite cycle of
packets, for example.  The second failure mode would cause packets to
expire prematurely.

The decapsulating node can implement a simple algorithm to prevent the
"gaining hop count" problem.  This algorithm does not prevent the second
problem.  This algorithm is implemented as part of the process of
decapsulating the IPv6 packet:

   -    If the tunnel is configured to use the "single-hop" model, do
        not modify the IPv6 hop limit field.

   -    If the tunnel is configured to use the "multi-hop" model, then:

           -    If the IPv4 TTL field is greater than or equal to the
                IPv6 hop limit field, do not modify the IPv6 hop limit
                field.

           -    Else, copy the IPv4 TTL field into the IPv6 hop limit
                field.

It is an open issue whether the "loosing too much hop count" problem is
serious enough to require that a solution be developed.

Note that the decision about whether to copy the IPv4 TTL field into the
hop limit field does not affect the requirement to decrement the hop
limit field; If the encapsulating or decapsulating node is an IPv6
router that forwards the packet, it must decrement the IPv6 hop count.

Note also that the hop limit problem affects only configured tunnels.
Automatic tunnels terminate at the end node, where the packet is
consumed, not forwarded, so the remaining hop limit is irrelevant.

4.1.3. Handling IPv4 ICMP errors

The encapsulating node has to be able to handle IPv4 ICMP errors that
are generated by routers interior to the tunnel. All such errors are
returned to the encapsulating node since the encapsulating node is the
IPv4 source of the packets.

Ideally the encapsulating node would want to convert these errors to
IPv6 ICMP errors and send them back to the source of the original IPv6
datagram.  However, this in infeasible since the IPv4 ICMP errors may
not return enough of the "offending packet".  Many IPv4 implementations
only return the IPv4 header plus 8 bytes of the IPv4 payload, which will
not even contain the complete IPv6 header, let alone enough higher level
headers for the originating node to determine which application
originated the packet that experienced the error.

For the purpose of this discussion there are two categories of errors:

  1)    ICMP errors that are needed to maintain connectivity.  Only ICMP
        "packet too big" falls in this category; a persistent loss of
        ICMP "packet too big" message would result in a black hole for
        large packets.

  2)    ICMP errors that are needed by network management tools like
        traceroute. These errors include ICMP unreachable and ICMP TTL
        expired.

The ICMP "packet too big" errors are handled according to IPv4 Path MTU
Discovery [8] and the resulting path MTU is recorded in the IPv4 layer.
The recorded path MTU is used by IPv6 to determine if an IPv6 ICMP

"packet too big" error has to be generated as described in section
4.1.1.

The other errors can be handled as described in the remainder of this
section to make multi-hop tunnels be "traceroute detectable."  Making a
tunnel traceroute detectable is implemented by having the encapsulating
node maintain "soft state" information about the tunnel.  This state is
created based on the IPv4 ICMP errors that are received in response to
encapsulated packets.  When the encapsulating node prepares to send an
IPv6 packet into a tunnel, it consults the tunnel state to determine if
the packet is likely to generate an ICMP error inside the tunnel.  If
so, it generates an appropriate IPv6 ICMP error, which it sends back to
the source of the IPv6 packet.  It also encapsulates the packet and
sends it into the tunnel. The latter is needed to quickly recover from
transient error conditions.

Note that, since the IPv6 ICMP error message originates at the
encapsulating node, not at the IPv4 router within the tunnel, the node
that sent the original IPv6 packet does not receive the address of the
IPv4 router.  Thus a traceroute program may not determine the addresses
of the IPv4 routers within a tunnel, but it may detect their presence by
noting that a packets with a consecutive range of hop limits expire at
the same router (the encapsulating router).

Tunnel state information is associated with the IPv4 address of the
endpoint of the tunnel and can include:

  -     The MTU of the Tunnel. Its use is described in section 4.1.1.

  -     Reachability of the endpoint of the tunnel.

  -     If the endpoint of the tunnel is unreachable, the IPv4 address
        of the router reporting unreachability.

  -     Path length of the tunnel (number of IPv4 hops to the endpoint).

  -     For each TTL 't' between 1 and the path length of the tunnel,
        the IPv4 address of the router that was last known to be 't'
        hops into the tunnel.

Maintaining the IPv4 addresses of the routers internal to the tunnel is
not strictly necessary for correct operation, but is useful for network
management.

The tunnel state does not have to be allocated until an IPv4 TTL.

4.1.3. Handling IPv4 ICMP error is
received. errors

In the absence of tunnel state, the tunnel MTU can be assumed response to be the MTU of the outgoing interface, the path length one hop and encapsulated packets it has sent into the
endpoint being reachable.

When tunnel, the
encapsulating node receives an may receive IPv4 ICMP error where the
"offending packet" is an IPv6-in-IPv4 packet (i.e. an messages from IPv4 packet with
an IP protocol field of 41),
routers inside the tunnel.  These packets are addressed to the
encapsulating node updates the tunnel
state associated with because it is the IPv4 destination in the "offending packet".
The update depends on the type of ICMP error:

   -    Host or network unreachable: Mark the tunnel endpoint as
        unreachable and record the source of the ICMP error as the
        source of unreachability.

   -    Time exceeded in transit: encapsulated
packet.

The TTL "consumed" before reaching the
        router that sent the time exceeded message is extracted from the
        IPv6 hop limit field in the "offending packet" (the IPv6 hop
        limit field is in the first 8 bytes of the IPv6 header thus it
        will be returned in the ICMP packet). Compute the updated tunnel
        path length as the maximum of the currently recorded path length
        and the extracted IPv6 hop limit. Record the source of the ICMP
        error as the router at 'IPv6 hop limit' hops into the tunnel.

   -    "Packet "packet too big": Use the big" error messages are handled according to IPv4
Path MTU Discovery [8]
        algorithm to update the tunnel MTU.

   -    For all other ICMP errors log a network management event.

When the encapsulating node prepares to forward an IPv6 packet into the
tunnel it performs the following checks against the tunnel state:

   -    If the tunnel endpoint is unreachable, it generates an IPv6 ICMP
        "destination unreachable" message.

   -    If and the hop limit resulting path MTU is less than the recorded tunnel TTL, it
        generates an IPv6 ICMP "time exceeded" message.

   -    If the packet would violate in the tunnel MTU, generate
IPv4 layer.  The recorded path MTU is used by IPv6 to determine if an
IPv6 ICMP "packet too big" message, error has to be generated as specified described in
section 4.1.1.

The IPv6 handling of other types of ICMP error message messages depends on how much
information is sent back to the source of included in the IPv6
packet, and includes as much of "packet in error" field, which holds the original IPv6
encapsulated packet as will fit.
The source IPv6 address that caused the error.

Many older IPv4 routers return only 8 bytes of data beyond the ICMP message is that IPv4
header of the encapsulating
node.  That original IPv6 packet in error, which is also forwarded into not enough to include the tunnel.

The algorithm as described above quickly returns IPv6 ICMP errors as a
result
address fields of the IPv6 header. More modern IPv4 ICMP errors from inside routers may return
enough data beyond the tunnel. In order IPv4 header to determine
when include the error condition is lifted, it relies on:

  -     A timeout.  All tunnel state, except entire IPv6 header and
possibly even the tunnel MTU, should be
        discarded after at most 30 seconds after it was created. data beyond that.

If the
        error condition still exists offending packet includes enough data, the encapsulating node may

extract the encapsulated IPv6 packet and packets continue use it to flow
        through that tunnel, IPv4 generating an IPv6
ICMP errors will continue message directed back to arrive
        and they will cause a refresh of the tunnel state.

        The tunnel MTU is timed out originating IPv6 node, as described in shown below:

                +--------------+
                | IPv4 Path MTU
        Discovery [8]. Header  |
                | dst = encaps |
                |       node   |
                +--------------+
                |     ICMP     |
                |    Header    |
         -     Data packets are always sent into the tunnel, even when the
        encapsulating -    +--------------+
                | IPv4 Header  |
                | src = encaps |
        IPv4    |       node generates   |
                +--------------+   - -
        Packet  |    IPv6      |
                |    Header    |   Original IPv6
         in     +--------------+   Packet -
                |  Transport   |   Can be used to
        Error   |    Header    |   generate an
                +--------------+   IPv6 ICMP error message.  This
        means that packets will get through as soon as the ICMP
                |              |   error
        condition within message
                ~     Data     ~   back to the tunnel is relieved, although error reports
        may continue for a short period thereafter. source.
                |              |
         - -    +--------------+   - -

        IPv4 ICMP Error Message Returned to Encapsulating Node

4.1.4.  IPv4 Header Construction

When encapsulating an IPv6 packet in an IPv4 datagram, the IPv4 header
fields are set as follows:

        Version:

                4

        IP Header Length in 32-bit words:

                5 (There are no IPv4 options in the encapsulating
                header.)

        Type of Service:

                0

        Total Length:

                Payload length from IPv6 header plus length of IPv6 and
                IPv4 headers (i.e. a constant 60 bytes).

        Identification:

                Generated uniquely as for any IPv4 packet transmitted by
                the system.

        Flags:

                Set the Don't Fragment (DF) flag as specified in
                section 4.1.1. Set the More Fragments (MF) bit as
                necessary if fragmenting.

        Fragment offset:

                Set as necessary if fragmenting.

        Time to Live:

                If tunnel is configured as multi-hop:

                        Copied from the IPv6 hop limit field.

                If tunnel is configured as single-hop:

                Set to pre-configured value. in implementation-specific manner.

        Protocol:

                41 (Assigned payload type number for IPv6)

        Header Checksum:

                Calculate the checksum of the IPv4 header.

        Source Address:

                IPv4 address of outgoing interface of the
                encapsulating node.

        Destination Address:

                IPv4 address of of tunnel endpoint.

Any IPv6 options are preserved in the packet (after the IPv6 header).

4.1.5. Decapsulating IPv6-in-IPv4 Packets

When an IPv6/IPv4 host or a router receives an IPv4 datagram that is
addressed to one of its own IPv4 address, and the value of the protocol
field is 41, it removes the IPv4 header and submits the IPv6 datagram to
its IPv6 layer code.

The decapsulation is shown below:

        +-------------+
        |    IPv4     |
        |   Header    |
        +-------------+                 +-------------+
        |    IPv6     |                 |    IPv6     |
        |   Header    |                 |   Header    |
        +-------------+                 +-------------+
        |  Transport  |                 |  Transport  |
        |   Layer     |      ===>       |   Layer     |
        |   Header    |                 |   Header    |
        +-------------+                 +-------------+
        |             |                 |             |
        ~    Data     ~                 ~    Data     ~
        |             |                 |             |
        +-------------+                 +-------------+

                    Decapsulating IPv6 from IPv4

When decapsulating the IPv6-in-IPv4 packet, only the hop limit field of the IPv6 header is modified:

        If tunnel is configured as single-hop:

                Do not modify the IPv6 hop limit field.

        If tunnel is configured as multi-hop:
modified.  If the IPv4 TTL field packet is greater than or equal to the
                IPv6 hop limit field, do not modify the IPv6 hop limit
                field.  Else, copy the IPv4 TTL field into the IPv6 subsequently forwarded, its hop limit field.

Then the is
decremented by one.

The encapsulating IPv4 header is discarded.

Note that the

The decapsulating node performs IPv4 reassembly before decapsulating the
IPv6 packet.  All IPv6 options are preserved even if the encapsulated encapsulating
IPv4 packet is fragmented.

After the IPv6 packet is decapsulated, it is treated processed the same as any
received IPv6 packet.

4.2. Configured Tunneling

In configured tunneling, the tunnel endpoint address is determined from
configuration information in the encapsulating node.  For each tunnel,
the encapsulating node must store the tunnel endpoint address.  When an
IPv6 packet is transmitted over a tunnel, the tunnel endpoint address
configured for that tunnel is used as the destination address for the
encapsulating IPv4 header.

The determination of which packets to tunnel is usually made by routing
information on the encapsulating node.  This is usually done via a
routing table, which directs packets based on their destination address
using the prefix mask and match technique.

4.2.1. Default Configured Tunnel

Nodes that are connected to IPv4 routing infrastructures may use a
configured tunnel to reach an IPv6 "backbone".  If the IPv4 address of
an IPv6/IPv4 router bordering the backbone is known, a tunnel can be
configured to that router.  This tunnel can be configured into the
routing table as a "default route".  That is, all destinations IPv6 destination
addresses will match the route and could potentially traverse the
tunnel.  Since the "mask length" of such default route is zero, it will
be used only if there are no other routes with a longer mask that match
the destination.

The tunnel endpoint address of such a default tunnel could be the IPv4
address of one IPv6/IPv4 router at the border of the IPv6 backbone.
Alternatively, the tunnel endpoint could be an IPv4 "logical "anycast address".
With this approach, multiple IPv6/IPv4 routers at the border advertise
IPv4 reachability to the same IPv4 logical address.  All of these routers accept
packets to this address as their own, and will decapsulate IPv6 packets
tunneled to this address.  This logical address
operates something like an "anycast address":  When an IPv6/IPv4 node
send sends an encapsulated
packet to this address, it will be delivered to only one of the border
routers, but the sending node will not know which one.  The IPv4 routing
system will generally carry the traffic to the closest router.

Using a default tunnel to a logical an IPv4 address "anycast address" provides a high
degree of robustness since multiple border router can be provided, and and,
using the normal fallback mechanisms of IPv4 routing, traffic will
automatically switch to another router when one goes down.

4.3. Automatic Tunneling

In automatic tunneling, the tunnel endpoint address is determined from
the packet being tunneled.  The destination IPv6 address in the packet
must be an IPv4-compatible address.  If it is, the IPv4 address
component of that address -- the low-order 32-bits -- are extracted and
used as the tunnel endpoint address.  IPv6 packets that are not
addressed to an IPv4-compatible address can not be tunneled using
automatic tunneling.

The determination of

IPv6/IPv4 nodes need to determine which IPv6 packets to automatically tunnel can be made
by sent via
automatic tunneling.  One technique is to use the IPv6 routing table information.  This to
direct automatic tunneling.  An implementation can be configured in the have a special static
routing table as route to entry for the prefix 0:0:0:0:0:0:0:0/96.  That 0:0:0:0:0:0/96.  (That is, a route to
the all-zeros prefix with a 96-bit mask. mask.)  Packets to all destinations
bearing the all-zeros 96-bit that match this
prefix can be are sent via to a pseudo-interface driver which performs automatic
tunneling.  Since all IPv4-compatible IPv6 addresses will match this
prefix, all packets to those destinations will be auto-tunneled.

4.4. Default Sending Algorithm

This section presents a combined IPv4 and IPv6 sending algorithm that
IPv6/IPv4 nodes can use.  The algorithm can be used to determine when to
send IPv4 packets, when to send IPv6 packets, and when to perform
automatic and configured tunneling.  It illustrates how the techniques
of dual IP layer, configured tunneling, and automatic tunneling can be
used together.  The algorithm has the following properties:

   -    Sends IPv4 packets to all IPv4 destinations.

   -    Sends IPv6 packets to all IPv6 destinations on the same link.

   -    Using automatic tunneling, sends IPv6 packets encapsulated in
        IPv4 to IPv6 destinations with IPv4-compatible addresses that
        are located off-link.

   -    Sends IPv6 packets to IPv6 destinations located off-link when
        IPv6 routers are present.

   -    Using the default IPv6 tunnel, sends IPv6 packets encapsulated
        in IPv4 to IPv6 destinations with IPv6-only addresses when no
        IPv6 routers are present.

The algorithm is as follows:

  1)    If the address of the end node is an IPv4 address then:

          1.1)  If the destination is located on the attached link, then
                send an IPv4 packet addressed to the end node.

          1.2)  If the destination is located off-link, then;

                1.2.1)  If there is an IPv4 router on link, then send an
                        IPv4 format packet.  The IPv4 destination
                        address is the IPv4 address of the end node.
                        The datalink address is the datalink address of
                        the IPv4 router.

                1.2.2)  Else, the destination is treated as
                        "unreachable" because it is located off link and
                        there are no on-link routers.

  2)    If the address of the end node is an IPv4-compatible IPv6
        address (i.e. bears the prefix 0:0:0:0:0:0), then:

          2.1)  If the destination is located on the attached link, then
                send an IPv6 format packet (not encapsulated).  The IPv6
                destination address is the IPv6 address of the end node.
                The datalink address is the datalink address of the end
                node.

          2.2)  If the destination is located off-link, then:

                2.2.1)  If there is an IPv4 router on the attached link,
                        then send an IPv6 packet encapsulated in IPv4.
                        The IPv6 destination address is the address of
                        the end node.  The IPv4 destination address is
                        the low-order 32-bits of the end node's address.
                        The datalink address is the datalink address of
                        the IPv4 router.

                2.2.2)  Else, if there is an IPv6 router on the attached
                        link, then send an IPv6 format packet.  The IPv6
                        destination address is the IPv6 address of the
                        end node.  The datalink address is the datalink
                        address of the IPv6 router.

                2.2.3)  Else, the destination is treated as
                        "unreachable" because it is located off-link and
                        there are no on-link routers.

   3)   If the address of the end node is an IPv6-only address, then:

          3.1)  If the destination is located on the attached link, then
                send an IPv6 format packet.  The IPv6 destination
                address is the IPv6 address of the end node.  The
                datalink address is the datalink address of the end
                node.

          3.2)  If the destination is located off-link, then:

                2.2.1)  If there is an IPv6 router on the attached link,
                        then send an IPv6 format packet.  The IPv6
                        destination address is the IPv6 address of the
                        end node.  The datalink address is the datalink
                        address of the IPv6 router.

                2.2.2)  Else, if the destination is reachable via a
                        configured tunnel, and there is an IPv4 router
                        on the attached link link, then send an IPv6
                        packet encapsulated in IPv4.  The IPv6
                        destination address is the address of the end
                        node.  The IPv4 destination address is the
                        configured IPv4 address of the tunnel endpoint.
                        The datalink address is the datalink address of
                        the IPv4 router.

                2.2.3)  Else, the destination is treated as
                        "unreachable" because it is located off-link and
                        there are no on-link IPv6 routers.

A summary of these sending rules are given in the table below:

End         | End     | IPv4    | IPv6    | Packet |      |      |
Node        | Node    | Router  | Router  | Format | IPv6 | IPv4 | DLink
Address     | On      | On      | On      | To     | Dest | Dest | Dest
Type        | Link?   | Link?   | Link?   | Send   | Addr | Addr | Addr
------------+---------+---------+---------+--------+------+------+------
IPv4        | Yes     |  N/A    |  N/A    | IPv4   |  N/A |  E4  | EL
------------+---------+---------+---------+--------+------+------+------
IPv4        | No      |  Yes    |  N/A    | IPv4   |  N/A |  E4  | RL
------------+---------+---------+---------+--------+------+------+------
IPv4        | No      |  No     |  N/A    | UNRCH  |  N/A |  N/A | N/A
------------+---------+---------+---------+--------+------+------+------
IPv4-compat | Yes     |  N/A    |  N/A    | IPv6   |  E6  |  N/A | EL
------------+---------+---------+---------+--------+------+------+------
IPv4-compat | No      |  Yes    |  N/A    | IPv6/4 |  E6  |  E4  | RL
------------+---------+---------+---------+--------+------+------+------
IPv4-compat | No      |  No     |  Yes    | IPv6   |  E6  |  N/A | RL
------------+---------+---------+---------+--------+------+------+------
IPv4-compat | No      |  No     |  No     | UNRCH  |  N/A |  N/A | N/A
------------+---------+---------+---------+--------+------+------+------
IPv6-only   | Yes     |  N/A    |  N/A    | IPv6   |  E6  |  N/A | EL
------------+---------+---------+---------+--------+------+------+------
IPv6-only   | No      |  N/A    |  Yes    | IPv6   |  E6  |  N/A | RL
------------+---------+---------+---------+--------+------+------+------
IPv6-only   | No      |  Yes    |  No     | IPv6/4 |  E6  |  T4  | RL
------------+---------+---------+---------+--------+------+------+------
IPv6-only   | No      |  No     |  No     | UNRCH  |  N/A |  N/A | N/A
------------+---------+---------+---------+--------+------+------+------

        Key to Abbreviations
        --------------------
        N/A:    Not applicable or does not matter.
        E6:     IPv6 address of end node.
        E4:     IPv4 address of end node (low-order 32-bits of
                IPv4-compatible address).
        EL:     Datalink address of end node.
        T4:     IPv4 address of the tunnel endpoint.
        R6:     IPv6 address of router.
        R4:     IPv4 address of router.
        RL:     Datalink address of router.
        IPv4:   IPv4 packet format.
        IPv6:   IPv6 packet format.
        IPv6/4: IPv6 encapsulated in IPv4 packet format.
        UNRCH:  Destination is unreachable.  Don't send a packet.

4.4.1  On/Off Link Determination

Part of the process of determining what packet format to use includes
determining whether a destination is located on an attached link or not.

IPv4 and IPv6 employ different mechanisms.  IPv4 uses an algorithm in
which the destination address and the interface address are both
logically ANDed with the netmask of the interface and then compared.  If
the resulting two values match, then the destination is located on-link.
This algorithm is discussed in more detail in Section 3.3.1.1 of the document "Requirements for Internet Hosts --
Communications Layers"
host requirements specification [11].  IPv6 uses the neighbor discovery
algorithm described in "IPv6 Neighbor Discovery -- Processing" [7].

IPv6/IPv4 nodes need to use both methods:

   -    If a destination is an IPv4 address, then the on/off link
        determination is made by comparison with the netmask, as
        described in RFC 1122 section 3.3.1.1.

   -    If a destination is represented by an IPv4-compatible IPv6
        address (prefix 0:0:0:0:0:0), the decision is made using the
        IPv4 netmask comparison algorithm using the low-order 32-bits
        (IPv4 address part) of the destination address.

  -     If the destination is represented by an IPv6-only address
        (prefix other than 0:0:0:0:0:0), the on/off link determination
        is made using the IPv6 neighbor discovery mechanism.

5. Acknowledgements

We would like to thank the members of the IPng working group and the
IPng transition working group for their many contributions and extensive
review of this document.

The  Special thanks to Jim Bound, Ross Callon, and
Bob Hinden for many helpful suggestions and to John Moy for suggesting
the IPv4 "logical "anycast address" default tunnel technique was originally
suggested by John Moy. technique.

6. Authors' Address

        Robert E. Gilligan
        Sun Microsystems, Inc.
        2550 Garcia Ave.
        Mailstop UMTV 05-44
        Mountain View, California 94043

        415-336-1012 (voice)
        415-336-6015 (fax)

        Bob.Gilligan@Eng.Sun.COM

        Erik Nordmark
        Sun Microsystems, Inc.
        2550 Garcia Ave.
        Mailstop UMTV 05-44
        Mountain View, California 94043

        415-336-2788 (voice)
        415-336-6015 (fax)

        Erik.Nordmark@Eng.Sun.COM

7. References

[1]     W. Croft, J. Gilmore.  "Bootstrap Protocol".  RFC 951.
        September 1985.

[2]     R. Droms.  "Dynamic Host Configuration Protocol".  RFC 1541.
        October 1993.

[3]     J. Bound, Y. Rekhter, Sue Thompson. "Dynamic Host Configuration
        Protocol for IPv6". Internet Draft
        <draft-ietf-dhc-dhcpv6-00.txt>. February 1995.

[4]     S. Deering, R. Hinden. "Internet Protocol, Version 6 (IPv6)
        Specification".  Internet Draft
        <draft-hinden-ipng-ipv6-spec-00.txt>.  October 1994.
        <draft-ietf-ipngwg-ipv6-spec-01.txt>, March 1995.

[5]     S. Thompson, IPv6 Stateless Address Configuration. Autoconfiguration, Internet
        Draft to be written. <draft-ietf-addrconf-ipv6-auto-01.txt>, March 1995.

[6]     S. Thompson, C. Huitema. "DNS Extensions to support IP version
        6".  Internet Draft <draft-thomson-ipng-dns-00.txt>.  October
        1994. <draft-ietf-ipngwg-dns-00.txt>, March 1995.

[7]     W. A. Simpson. "IPv6 Neighbor Discovery -- Processing".
        Internet Draft <draft-simpson-ipv6-discov-process-00.txt>.
        October 1994. <draft-simpson-ipv6-discov-process-02.txt>.
        February 1995.

[8]     J. Mogul, S. Deering.  "Path MTU Discovery". RFC 1191.  November
        1990.

[9]     R. Finlayson, T. Mann, J. Mogul, M. Theimer. "Reverse Address
        Resolution Protocol".  RFC 903.  June 1984.

[10]    R. Braden. "Requirements for Internet Hosts - Application And
        Support". RFC 1123. October 1989.

[11]    R. Braden. "Requirements for Internet Hosts - Communication
        Layers". RFC 1122. October 1989.

[12]    A. Conta, S. Deering. "ICMP for the Internet Protocol Version 6
        (IPv6)". Internet Draft <draft-ietf-ipngwg-icmp-01.txt>.
        February 1995.

[13]    C. Kent and J. Mogul.  "Fragmentation Considered Harmful".  In
        Proc.  SIGCOMM '87 Workshop on Frontiers in Computer
        Communications Technology.  August, 1987.